convert_qat_model_for_ptq
-utils.quantization.convert_qat_model_for_ptq(model, *, quantize_embedding=None)This function is used to convert a swap fake-quantized modules in a model -which has been trained with QAT back to the original modules, ready for PTQ.
-Parameters
-| Name | -Type | -Description | -Default | -||
|---|---|---|---|---|---|
| model | -- | The model to convert. | -required | -||
| quantize_embedding | -bool | None | -Whether to quantize the model’s embedding weights. | -None |
+quantize_model | +This function is used to quantize a model. |
convert_qat_model
+utils.quantization.convert_qat_model(model, quantize_embedding=False)This function converts a QAT model which has fake quantized layers back to the original model.
get_ptq_config
-utils.quantization.get_ptq_config(
+
+get_quantization_config
+utils.quantization.get_quantization_config(
weight_dtype,
activation_dtype=None,
group_size=None,
)
This function is used to build a post-training quantization config.
-
-Parameters
+
+Parameters
@@ -615,13 +581,13 @@ which has been trained with QAT back to the original modules, ready for PTQ.
weight_dtype
-TorchIntDType
+TorchAOQuantDType
The dtype to use for weight quantization.
required
activation_dtype
-TorchIntDType | None
+TorchAOQuantDType | None
The dtype to use for activation quantization.
None
@@ -683,21 +649,21 @@ which has been trained with QAT back to the original modules, ready for PTQ.
utils.quantization.prepare_model_for_qat(
model,
weight_dtype,
- group_size,
+ group_size=None,
activation_dtype=None,
quantize_embedding=False,
)
This function is used to prepare a model for QAT by swapping the model’s linear
layers with fake quantized linear layers, and optionally the embedding weights with
fake quantized embedding weights.
-
-Parameters
+
+Parameters
@@ -716,19 +682,19 @@ fake quantized embedding weights.
weight_dtype
-TorchIntDType
+TorchAOQuantDType
The dtype to use for weight quantization.
required
group_size
-int
+int | None
The group size to use for weight quantization.
-required
+None
activation_dtype
-TorchIntDType | None
+TorchAOQuantDType | None
The dtype to use for activation quantization.
None
@@ -766,26 +732,24 @@ fake quantized embedding weights.
-
-quantize_model_for_ptq
-utils.quantization.quantize_model_for_ptq(
+
+quantize_model
+utils.quantization.quantize_model(
model,
weight_dtype,
group_size=None,
activation_dtype=None,
quantize_embedding=None,
)
-This function is used to quantize a model for post-training quantization.
-It swaps the model’s linear layers with fake quantized linear layers.
-If quantize_embedding is True, it will also swap the model’s embedding weights with fake quantized embedding weights.
-
-Parameters
+This function is used to quantize a model.
+
+Parameters
@@ -804,7 +768,7 @@ If quantize_embedding is True, it will also swap the model’s embe
weight_dtype
-TorchIntDType
+TorchAOQuantDType
The dtype to use for weight quantization.
required
@@ -816,7 +780,7 @@ If quantize_embedding is True, it will also swap the model’s embe
activation_dtype
-TorchIntDType | None
+TorchAOQuantDType | None
The dtype to use for activation quantization.
None
diff --git a/docs/api/utils.schemas.enums.html b/docs/api/utils.schemas.enums.html
index 90b478ab2..4c0906f02 100644
--- a/docs/api/utils.schemas.enums.html
+++ b/docs/api/utils.schemas.enums.html
@@ -505,7 +505,6 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
CustomSupportedOptimizers
RLType
RingAttnFunc
- TorchIntDType
@@ -547,10 +546,6 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
RingAttnFunc
Enum class for supported ring-flash-attn implementations
-
-TorchIntDType
-Torch integer data types - getattr guards against torch < 2.6 which does not support int4
-
@@ -572,11 +567,6 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
RingAttnFunc
utils.schemas.enums.RingAttnFunc()
Enum class for supported ring-flash-attn implementations
-
-
-TorchIntDType
-utils.schemas.enums.TorchIntDType()
-Torch integer data types - getattr guards against torch < 2.6 which does not support int4
diff --git a/docs/config-reference.html b/docs/config-reference.html
index dbe666e63..fe6b23a56 100644
--- a/docs/config-reference.html
+++ b/docs/config-reference.html
@@ -643,1230 +643,1226 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
qat: QATConfig | None
# For QATConfig:
- # Fake quantization layout to use for activation quantization. Valid options are
- # "int4" and "int8"
- activation_dtype: TorchIntDType | None
- # Fake quantization layout to use for weight quantization. Valid options are "int4"
- # and "int8"
- weight_dtype: TorchIntDType = TorchIntDType.int8
- # Quantize embedding
- quantize_embedding: bool | None = False
- # The number of elements in each group for per-group fake quantization
- group_size: int | None = 32
- # The number of steps to apply fake quantization after
- fake_quant_after_n_steps: int | None
-
-quantization: PTQConfig | None
- # For PTQConfig:
- # Fake quantization layout to use for weight quantization. Valid options are uintX for
- # X in [1, 2, 3, 4, 5, 6, 7], or int4, or int8
- weight_dtype: TorchIntDType = TorchIntDType.int8
- # Fake quantization layout to use for activation quantization. Valid options are
- # "int4" and "int8"
- activation_dtype: TorchIntDType | None
- # Whether to quantize the embedding layer.
- quantize_embedding: bool | None
- # The number of elements in each group for per-group fake quantization
- group_size: int | None = 32
-
-# Reward modelling: `True` or `False`
-reward_model: bool | None
-# Process reward modelling: `True` or `False`
-process_reward_model: bool | None
-# Coefficient to incentivize the reward model to output mean-zero rewards (proposed by
-# https://huggingface.co/papers/2312.09244, Eq. 2). Recommended value: `0.01`.
-center_rewards_coefficient: float | None
-num_labels: int | None
-
-# Whether to perform weighting in DPO trainer
-dpo_use_weighting: bool | None
-dpo_use_logits_to_keep: bool | None
-dpo_label_smoothing: float | None
-dpo_norm_loss: bool | None
-dpo_padding_free: bool | None
-dpo_generate_during_eval: bool | None
-
-# A list of one or more datasets to finetune the model with
-datasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None
- # For SFTDataset:
- # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory
- path: str | None
- # name of dataset split to load from
- split: str | None
- # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]
- type: str | UserDefinedPrompterType | None
- # For UserDefinedPrompterType:
- # Custom user instruction prompt
- system_prompt: str | None
- # Use {system} as key to be replaced
- system_format: str | None
- field_system: str | None
- field_instruction: str | None
- field_input: str | None
- field_output: str | None
-
- # Customizable to be single line or multi-line. Use {instruction}/{input} as key to
- # be replaced. 'format' can include {input}
- format: str | None
- # 'no_input_format' cannot include {input}
- no_input_format: str | None
- input_transform: str | None
- # split dataset into N pieces (use with shards_idx)
- shards: int | None
- # the index of sharded dataset to use
- shards_idx: int | None
- # process dataset in N sequential chunks for memory efficiency (exclusive with
- # `shards`)
- preprocess_shards: int | None
- conversation: str | None
-
- # The name of the chat template to use for training, following values are supported:
- # tokenizer_default: Uses the chat template that is available in the
- # tokenizer_config.json. If the chat template is not available in the tokenizer, it
- # will raise an error. This is the default.
- # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
- # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
- # tokenizer_default_fallback_*: where * is the name of the chat template to fallback
- # to if the tokenizer does not have a chat template else default to tokenizer. E.g.
- # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat
- # template. The custom jinja template should be provided in the chat_template_jinja
- # field.
- chat_template: ChatTemplate | str | None
- # Custom jinja chat template or path to jinja file. Used only if `chat_template:
- # jinja` or empty.
- chat_template_jinja: str | None
- # path to source data files
- data_files: str | list[str] | None
- input_format: str | None
- # name of dataset configuration to load
- name: str | None
- # defines the datatype when path is a file
- ds_type: str | None
- # For `completion` datasets only, uses the provided field instead of `text` column
- field: str | None
- field_human: str | None
- field_model: str | None
- # Key containing the messages (default: "messages")
- field_messages: str | None
- # Key containing the tools (default: "tools"). Must be a list[dict] and follow [JSON
- # schema](https://json-schema.org/learn/getting-started-step-by-step).
- field_tools: str | None
- # Key containing the reasoning trace (default: "reasoning_content").
- field_thinking: str | None
- # The key the chat template expects that indicates the reasoning trace.
- template_thinking_key: str | None
-
- message_field_role: str | None
-
- message_field_content: str | None
- # Mapping of properties from the input dataset to the chat template. (default:
- # message_property_mappings={'role':'role', 'content':'content'}) If a property exists
- # in the template but not in this mapping, the system will attempt to load it directly
- # from the message using the property name as the key. Example: In the mapping below,
- # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and
- # used as 'content' in the chat template.
- message_property_mappings: dict[str, str] | None
- # The key in the message turn that indicates via boolean whether tokens of a turn
- # should be considered for training. Useful to selectively train on certain turns
- # besides the `roles_to_train`.
- message_field_training: str | None
- # The key in the message turn that contains the training details. Useful to
- # selectively train on certain tokens in a turn. The value of the key is a List[Dict]
- # containing `begin_offset` (start character index in content), `end_offset` (end
- # character index in content), and `train` (boolean whether to train).
- message_field_training_detail: str | None
- # (for Qwen3 template only) Whether to split the assistant content based on a
- # reasoning trace inside delimited tags
- split_thinking: bool | None
- logprobs_field: str | None
- temperature: float | None
- # Roles to train on. The tokens from these roles will be considered for the loss.
- roles_to_train: list[str] | None
- # Which EOS tokens to train on in the conversation. Possible values are: all: train on
- # all EOS tokens, turn (default): train on the EOS token at the end of each trainable
- # turn, last: train on the last EOS token in the conversation
- train_on_eos: Literal['all', 'turn', 'last'] | None
- # Roles mapping in the messages. The format is {target_role: [source_roles]}. All
- # source roles will be mapped to the target role. The default is: user: ["human",
- # "user"], assistant: ["gpt", "assistant"], system: ["system"], tool: ["tool"]
- roles: dict[str, list[str]] | None
- # Whether to drop the system turn from the dataset. Only works with chat_template.
- # This does not drop the default system message from chat_template if it exists. If
- # you wish to, we recommend using a custom jinja template with the default system
- # message removed or adding a system turn with empty content.
- drop_system_message: bool | None
- # Trust remote code for untrusted source
- trust_remote_code: bool | None = False
- # The specific revision of the dataset to use when loading from the Hugging Face Hub.
- # This can be a commit hash, tag, or branch name. If not specified, the latest version
- # will be used. This parameter is ignored for local datasets.
- revision: str | None
-
- # For DPODataset:
- path: str | None
- split: str | None
- type: UserDefinedDPOType | str | None
- # For UserDefinedDPOType:
- field_system: str | None
- field_prompt: str | None
- field_chosen: str | None
- field_rejected: str | None
- prompt_format: str | None
- chosen_format: str | None
- rejected_format: str | None
- data_files: list[str] | None
- revision: str | None
- field_messages: str | None
-
- # For KTODataset:
- path: str | None
- split: str | None
- type: UserDefinedKTOType | str | None
- # For UserDefinedKTOType:
- field_system: str | None
- field_prompt: str | None
- field_completion: str | None
- field_label: bool | None
- prompt_format: str | None
- completion_format: str | None
- data_files: list[str] | None
- trust_remote_code: bool | None = False
- revision: str | None
-
- # For StepwiseSupervisedDataset:
- path: str | None
- split: str | None
- data_files: list[str] | None
- revision: str | None
- step_separator: str | None
- max_completion_length: int | None
- train_on_last_step_only: bool | None
-
-# A list of one or more datasets to eval the model with. You can use either
-# test_datasets, or val_set_size, but not both.
-test_datasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None
- # For SFTDataset:
- # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory
- path: str | None
- # name of dataset split to load from
- split: str | None
- # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]
- type: str | UserDefinedPrompterType | None
- # For UserDefinedPrompterType:
- # Custom user instruction prompt
- system_prompt: str | None
- # Use {system} as key to be replaced
- system_format: str | None
- field_system: str | None
- field_instruction: str | None
- field_input: str | None
- field_output: str | None
-
- # Customizable to be single line or multi-line. Use {instruction}/{input} as key to
- # be replaced. 'format' can include {input}
- format: str | None
- # 'no_input_format' cannot include {input}
- no_input_format: str | None
- input_transform: str | None
- # split dataset into N pieces (use with shards_idx)
- shards: int | None
- # the index of sharded dataset to use
- shards_idx: int | None
- # process dataset in N sequential chunks for memory efficiency (exclusive with
- # `shards`)
- preprocess_shards: int | None
- conversation: str | None
-
- # The name of the chat template to use for training, following values are supported:
- # tokenizer_default: Uses the chat template that is available in the
- # tokenizer_config.json. If the chat template is not available in the tokenizer, it
- # will raise an error. This is the default.
- # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
- # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
- # tokenizer_default_fallback_*: where * is the name of the chat template to fallback
- # to if the tokenizer does not have a chat template else default to tokenizer. E.g.
- # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat
- # template. The custom jinja template should be provided in the chat_template_jinja
- # field.
- chat_template: ChatTemplate | str | None
- # Custom jinja chat template or path to jinja file. Used only if `chat_template:
- # jinja` or empty.
- chat_template_jinja: str | None
- # path to source data files
- data_files: str | list[str] | None
- input_format: str | None
- # name of dataset configuration to load
- name: str | None
- # defines the datatype when path is a file
- ds_type: str | None
- # For `completion` datasets only, uses the provided field instead of `text` column
- field: str | None
- field_human: str | None
- field_model: str | None
- # Key containing the messages (default: "messages")
- field_messages: str | None
- # Key containing the tools (default: "tools"). Must be a list[dict] and follow [JSON
- # schema](https://json-schema.org/learn/getting-started-step-by-step).
- field_tools: str | None
- # Key containing the reasoning trace (default: "reasoning_content").
- field_thinking: str | None
- # The key the chat template expects that indicates the reasoning trace.
- template_thinking_key: str | None
-
- message_field_role: str | None
-
- message_field_content: str | None
- # Mapping of properties from the input dataset to the chat template. (default:
- # message_property_mappings={'role':'role', 'content':'content'}) If a property exists
- # in the template but not in this mapping, the system will attempt to load it directly
- # from the message using the property name as the key. Example: In the mapping below,
- # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and
- # used as 'content' in the chat template.
- message_property_mappings: dict[str, str] | None
- # The key in the message turn that indicates via boolean whether tokens of a turn
- # should be considered for training. Useful to selectively train on certain turns
- # besides the `roles_to_train`.
- message_field_training: str | None
- # The key in the message turn that contains the training details. Useful to
- # selectively train on certain tokens in a turn. The value of the key is a List[Dict]
- # containing `begin_offset` (start character index in content), `end_offset` (end
- # character index in content), and `train` (boolean whether to train).
- message_field_training_detail: str | None
- # (for Qwen3 template only) Whether to split the assistant content based on a
- # reasoning trace inside delimited tags
- split_thinking: bool | None
- logprobs_field: str | None
- temperature: float | None
- # Roles to train on. The tokens from these roles will be considered for the loss.
- roles_to_train: list[str] | None
- # Which EOS tokens to train on in the conversation. Possible values are: all: train on
- # all EOS tokens, turn (default): train on the EOS token at the end of each trainable
- # turn, last: train on the last EOS token in the conversation
- train_on_eos: Literal['all', 'turn', 'last'] | None
- # Roles mapping in the messages. The format is {target_role: [source_roles]}. All
- # source roles will be mapped to the target role. The default is: user: ["human",
- # "user"], assistant: ["gpt", "assistant"], system: ["system"], tool: ["tool"]
- roles: dict[str, list[str]] | None
- # Whether to drop the system turn from the dataset. Only works with chat_template.
- # This does not drop the default system message from chat_template if it exists. If
- # you wish to, we recommend using a custom jinja template with the default system
- # message removed or adding a system turn with empty content.
- drop_system_message: bool | None
- # Trust remote code for untrusted source
- trust_remote_code: bool | None = False
- # The specific revision of the dataset to use when loading from the Hugging Face Hub.
- # This can be a commit hash, tag, or branch name. If not specified, the latest version
- # will be used. This parameter is ignored for local datasets.
- revision: str | None
-
- # For DPODataset:
- path: str | None
- split: str | None
- type: UserDefinedDPOType | str | None
- # For UserDefinedDPOType:
- field_system: str | None
- field_prompt: str | None
- field_chosen: str | None
- field_rejected: str | None
- prompt_format: str | None
- chosen_format: str | None
- rejected_format: str | None
- data_files: list[str] | None
- revision: str | None
- field_messages: str | None
-
- # For KTODataset:
- path: str | None
- split: str | None
- type: UserDefinedKTOType | str | None
- # For UserDefinedKTOType:
- field_system: str | None
- field_prompt: str | None
- field_completion: str | None
- field_label: bool | None
- prompt_format: str | None
- completion_format: str | None
- data_files: list[str] | None
- trust_remote_code: bool | None = False
- revision: str | None
-
- # For StepwiseSupervisedDataset:
- path: str | None
- split: str | None
- data_files: list[str] | None
- revision: str | None
- step_separator: str | None
- max_completion_length: int | None
- train_on_last_step_only: bool | None
-
-# If false, the datasets will not be shuffled and will keep their original order in
-# `datasets`. The same applies to the `test_datasets` option and the
-# `pretraining_dataset` option. Default is true.
-shuffle_merged_datasets: bool | None = True
-# If true, each dataset in `datasets` will be shuffled before merging. This allows
-# curriculum learning strategies to be applied at the dataset level. Default is false.
-shuffle_before_merging_datasets: bool | None = False
-# Axolotl attempts to save the dataset as an arrow after packing the data together so
-# subsequent training attempts load faster, relative path
-dataset_prepared_path: str | None
-# Num shards for whole dataset
-dataset_shard_num: int | None
-# Index of shard to use for whole dataset
-dataset_shard_idx: int | None
-skip_prepare_dataset: bool | None = False
-# Number of shards to save the prepared dataset
-num_dataset_shards_to_save: int | None
-
-# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize
-pretraining_dataset: Annotated[list[PretrainingDataset | SFTDataset], MinLen(1)] | None
- # For PretrainingDataset:
- name: str | None
- path: str | None
- split: str | None = train
- text_column: str | None = text
- type: str | None = pretrain
- trust_remote_code: bool | None = False
- data_files: str | None
- skip: int | None
-
- # For SFTDataset:
- # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory
- path: str | None
- # name of dataset split to load from
- split: str | None
- # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]
- type: str | UserDefinedPrompterType | None
- # For UserDefinedPrompterType:
- # Custom user instruction prompt
- system_prompt: str | None
- # Use {system} as key to be replaced
- system_format: str | None
- field_system: str | None
- field_instruction: str | None
- field_input: str | None
- field_output: str | None
-
- # Customizable to be single line or multi-line. Use {instruction}/{input} as key to
- # be replaced. 'format' can include {input}
- format: str | None
- # 'no_input_format' cannot include {input}
- no_input_format: str | None
- input_transform: str | None
- # split dataset into N pieces (use with shards_idx)
- shards: int | None
- # the index of sharded dataset to use
- shards_idx: int | None
- # process dataset in N sequential chunks for memory efficiency (exclusive with
- # `shards`)
- preprocess_shards: int | None
- conversation: str | None
-
- # The name of the chat template to use for training, following values are supported:
- # tokenizer_default: Uses the chat template that is available in the
- # tokenizer_config.json. If the chat template is not available in the tokenizer, it
- # will raise an error. This is the default.
- # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
- # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
- # tokenizer_default_fallback_*: where * is the name of the chat template to fallback
- # to if the tokenizer does not have a chat template else default to tokenizer. E.g.
- # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat
- # template. The custom jinja template should be provided in the chat_template_jinja
- # field.
- chat_template: ChatTemplate | str | None
- # Custom jinja chat template or path to jinja file. Used only if `chat_template:
- # jinja` or empty.
- chat_template_jinja: str | None
- # path to source data files
- data_files: str | list[str] | None
- input_format: str | None
- # name of dataset configuration to load
- name: str | None
- # defines the datatype when path is a file
- ds_type: str | None
- # For `completion` datasets only, uses the provided field instead of `text` column
- field: str | None
- field_human: str | None
- field_model: str | None
- # Key containing the messages (default: "messages")
- field_messages: str | None
- # Key containing the tools (default: "tools"). Must be a list[dict] and follow [JSON
- # schema](https://json-schema.org/learn/getting-started-step-by-step).
- field_tools: str | None
- # Key containing the reasoning trace (default: "reasoning_content").
- field_thinking: str | None
- # The key the chat template expects that indicates the reasoning trace.
- template_thinking_key: str | None
-
- message_field_role: str | None
-
- message_field_content: str | None
- # Mapping of properties from the input dataset to the chat template. (default:
- # message_property_mappings={'role':'role', 'content':'content'}) If a property exists
- # in the template but not in this mapping, the system will attempt to load it directly
- # from the message using the property name as the key. Example: In the mapping below,
- # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and
- # used as 'content' in the chat template.
- message_property_mappings: dict[str, str] | None
- # The key in the message turn that indicates via boolean whether tokens of a turn
- # should be considered for training. Useful to selectively train on certain turns
- # besides the `roles_to_train`.
- message_field_training: str | None
- # The key in the message turn that contains the training details. Useful to
- # selectively train on certain tokens in a turn. The value of the key is a List[Dict]
- # containing `begin_offset` (start character index in content), `end_offset` (end
- # character index in content), and `train` (boolean whether to train).
- message_field_training_detail: str | None
- # (for Qwen3 template only) Whether to split the assistant content based on a
- # reasoning trace inside delimited tags
- split_thinking: bool | None
- logprobs_field: str | None
- temperature: float | None
- # Roles to train on. The tokens from these roles will be considered for the loss.
- roles_to_train: list[str] | None
- # Which EOS tokens to train on in the conversation. Possible values are: all: train on
- # all EOS tokens, turn (default): train on the EOS token at the end of each trainable
- # turn, last: train on the last EOS token in the conversation
- train_on_eos: Literal['all', 'turn', 'last'] | None
- # Roles mapping in the messages. The format is {target_role: [source_roles]}. All
- # source roles will be mapped to the target role. The default is: user: ["human",
- # "user"], assistant: ["gpt", "assistant"], system: ["system"], tool: ["tool"]
- roles: dict[str, list[str]] | None
- # Whether to drop the system turn from the dataset. Only works with chat_template.
- # This does not drop the default system message from chat_template if it exists. If
- # you wish to, we recommend using a custom jinja template with the default system
- # message removed or adding a system turn with empty content.
- drop_system_message: bool | None
- # Trust remote code for untrusted source
- trust_remote_code: bool | None = False
- # The specific revision of the dataset to use when loading from the Hugging Face Hub.
- # This can be a commit hash, tag, or branch name. If not specified, the latest version
- # will be used. This parameter is ignored for local datasets.
- revision: str | None
-
-# The maximum number of processes to use while preprocessing your input dataset. This
-# defaults to `os.cpu_count()` if not set. For Runpod VMs, it will default to number of
-# vCPUs via RUNPOD_CPU_COUNT.
-dataset_processes: int | None
-# Deduplicates datasets and test_datasets with identical entries
-dataset_exact_deduplication: bool | None
-# Keep dataset in memory while preprocessing. Only needed if cached dataset is taking
-# too much storage
-dataset_keep_in_memory: bool | None
-dataloader_pin_memory: bool | None
-dataloader_num_workers: int | None
-dataloader_prefetch_factor: int | None
-dataloader_drop_last: bool | None
+ # Fake quantization layout to use for activation quantization.
+ activation_dtype: TorchAOQuantDType | None
+ # Fake quantization layout to use for weight quantization.
+ weight_dtype: TorchAOQuantDType = TorchAOQuantDType.int8
+ # Quantize embedding
+ quantize_embedding: bool | None = False
+ # The number of elements in each group for per-group fake quantization
+ group_size: int | None = 32
+ # The number of steps to apply fake quantization after
+ fake_quant_after_n_steps: int | None
+
+quantization: PTQConfig | None
+ # For PTQConfig:
+ # Fake quantization layout to use for weight quantization.
+ weight_dtype: TorchAOQuantDType = TorchAOQuantDType.int8
+ # Fake quantization layout to use for activation quantization.
+ activation_dtype: TorchAOQuantDType | None
+ # Whether to quantize the embedding layer.
+ quantize_embedding: bool | None
+ # The number of elements in each group for per-group fake quantization
+ group_size: int | None = 32
+
+# Reward modelling: `True` or `False`
+reward_model: bool | None
+# Process reward modelling: `True` or `False`
+process_reward_model: bool | None
+# Coefficient to incentivize the reward model to output mean-zero rewards (proposed by
+# https://huggingface.co/papers/2312.09244, Eq. 2). Recommended value: `0.01`.
+center_rewards_coefficient: float | None
+num_labels: int | None
+
+# Whether to perform weighting in DPO trainer
+dpo_use_weighting: bool | None
+dpo_use_logits_to_keep: bool | None
+dpo_label_smoothing: float | None
+dpo_norm_loss: bool | None
+dpo_padding_free: bool | None
+dpo_generate_during_eval: bool | None
+
+# A list of one or more datasets to finetune the model with
+datasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None
+ # For SFTDataset:
+ # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory
+ path: str | None
+ # name of dataset split to load from
+ split: str | None
+ # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]
+ type: str | UserDefinedPrompterType | None
+ # For UserDefinedPrompterType:
+ # Custom user instruction prompt
+ system_prompt: str | None
+ # Use {system} as key to be replaced
+ system_format: str | None
+ field_system: str | None
+ field_instruction: str | None
+ field_input: str | None
+ field_output: str | None
+
+ # Customizable to be single line or multi-line. Use {instruction}/{input} as key to
+ # be replaced. 'format' can include {input}
+ format: str | None
+ # 'no_input_format' cannot include {input}
+ no_input_format: str | None
+ input_transform: str | None
+ # split dataset into N pieces (use with shards_idx)
+ shards: int | None
+ # the index of sharded dataset to use
+ shards_idx: int | None
+ # process dataset in N sequential chunks for memory efficiency (exclusive with
+ # `shards`)
+ preprocess_shards: int | None
+ conversation: str | None
+
+ # The name of the chat template to use for training, following values are supported:
+ # tokenizer_default: Uses the chat template that is available in the
+ # tokenizer_config.json. If the chat template is not available in the tokenizer, it
+ # will raise an error. This is the default.
+ # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
+ # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
+ # tokenizer_default_fallback_*: where * is the name of the chat template to fallback
+ # to if the tokenizer does not have a chat template else default to tokenizer. E.g.
+ # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat
+ # template. The custom jinja template should be provided in the chat_template_jinja
+ # field.
+ chat_template: ChatTemplate | str | None
+ # Custom jinja chat template or path to jinja file. Used only if `chat_template:
+ # jinja` or empty.
+ chat_template_jinja: str | None
+ # path to source data files
+ data_files: str | list[str] | None
+ input_format: str | None
+ # name of dataset configuration to load
+ name: str | None
+ # defines the datatype when path is a file
+ ds_type: str | None
+ # For `completion` datasets only, uses the provided field instead of `text` column
+ field: str | None
+ field_human: str | None
+ field_model: str | None
+ # Key containing the messages (default: "messages")
+ field_messages: str | None
+ # Key containing the tools (default: "tools"). Must be a list[dict] and follow [JSON
+ # schema](https://json-schema.org/learn/getting-started-step-by-step).
+ field_tools: str | None
+ # Key containing the reasoning trace (default: "reasoning_content").
+ field_thinking: str | None
+ # The key the chat template expects that indicates the reasoning trace.
+ template_thinking_key: str | None
+
+ message_field_role: str | None
+
+ message_field_content: str | None
+ # Mapping of properties from the input dataset to the chat template. (default:
+ # message_property_mappings={'role':'role', 'content':'content'}) If a property exists
+ # in the template but not in this mapping, the system will attempt to load it directly
+ # from the message using the property name as the key. Example: In the mapping below,
+ # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and
+ # used as 'content' in the chat template.
+ message_property_mappings: dict[str, str] | None
+ # The key in the message turn that indicates via boolean whether tokens of a turn
+ # should be considered for training. Useful to selectively train on certain turns
+ # besides the `roles_to_train`.
+ message_field_training: str | None
+ # The key in the message turn that contains the training details. Useful to
+ # selectively train on certain tokens in a turn. The value of the key is a List[Dict]
+ # containing `begin_offset` (start character index in content), `end_offset` (end
+ # character index in content), and `train` (boolean whether to train).
+ message_field_training_detail: str | None
+ # (for Qwen3 template only) Whether to split the assistant content based on a
+ # reasoning trace inside delimited tags
+ split_thinking: bool | None
+ logprobs_field: str | None
+ temperature: float | None
+ # Roles to train on. The tokens from these roles will be considered for the loss.
+ roles_to_train: list[str] | None
+ # Which EOS tokens to train on in the conversation. Possible values are: all: train on
+ # all EOS tokens, turn (default): train on the EOS token at the end of each trainable
+ # turn, last: train on the last EOS token in the conversation
+ train_on_eos: Literal['all', 'turn', 'last'] | None
+ # Roles mapping in the messages. The format is {target_role: [source_roles]}. All
+ # source roles will be mapped to the target role. The default is: user: ["human",
+ # "user"], assistant: ["gpt", "assistant"], system: ["system"], tool: ["tool"]
+ roles: dict[str, list[str]] | None
+ # Whether to drop the system turn from the dataset. Only works with chat_template.
+ # This does not drop the default system message from chat_template if it exists. If
+ # you wish to, we recommend using a custom jinja template with the default system
+ # message removed or adding a system turn with empty content.
+ drop_system_message: bool | None
+ # Trust remote code for untrusted source
+ trust_remote_code: bool | None = False
+ # The specific revision of the dataset to use when loading from the Hugging Face Hub.
+ # This can be a commit hash, tag, or branch name. If not specified, the latest version
+ # will be used. This parameter is ignored for local datasets.
+ revision: str | None
+
+ # For DPODataset:
+ path: str | None
+ split: str | None
+ type: UserDefinedDPOType | str | None
+ # For UserDefinedDPOType:
+ field_system: str | None
+ field_prompt: str | None
+ field_chosen: str | None
+ field_rejected: str | None
+ prompt_format: str | None
+ chosen_format: str | None
+ rejected_format: str | None
+ data_files: list[str] | None
+ revision: str | None
+ field_messages: str | None
+
+ # For KTODataset:
+ path: str | None
+ split: str | None
+ type: UserDefinedKTOType | str | None
+ # For UserDefinedKTOType:
+ field_system: str | None
+ field_prompt: str | None
+ field_completion: str | None
+ field_label: bool | None
+ prompt_format: str | None
+ completion_format: str | None
+ data_files: list[str] | None
+ trust_remote_code: bool | None = False
+ revision: str | None
+
+ # For StepwiseSupervisedDataset:
+ path: str | None
+ split: str | None
+ data_files: list[str] | None
+ revision: str | None
+ step_separator: str | None
+ max_completion_length: int | None
+ train_on_last_step_only: bool | None
+
+# A list of one or more datasets to eval the model with. You can use either
+# test_datasets, or val_set_size, but not both.
+test_datasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None
+ # For SFTDataset:
+ # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory
+ path: str | None
+ # name of dataset split to load from
+ split: str | None
+ # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]
+ type: str | UserDefinedPrompterType | None
+ # For UserDefinedPrompterType:
+ # Custom user instruction prompt
+ system_prompt: str | None
+ # Use {system} as key to be replaced
+ system_format: str | None
+ field_system: str | None
+ field_instruction: str | None
+ field_input: str | None
+ field_output: str | None
+
+ # Customizable to be single line or multi-line. Use {instruction}/{input} as key to
+ # be replaced. 'format' can include {input}
+ format: str | None
+ # 'no_input_format' cannot include {input}
+ no_input_format: str | None
+ input_transform: str | None
+ # split dataset into N pieces (use with shards_idx)
+ shards: int | None
+ # the index of sharded dataset to use
+ shards_idx: int | None
+ # process dataset in N sequential chunks for memory efficiency (exclusive with
+ # `shards`)
+ preprocess_shards: int | None
+ conversation: str | None
+
+ # The name of the chat template to use for training, following values are supported:
+ # tokenizer_default: Uses the chat template that is available in the
+ # tokenizer_config.json. If the chat template is not available in the tokenizer, it
+ # will raise an error. This is the default.
+ # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
+ # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
+ # tokenizer_default_fallback_*: where * is the name of the chat template to fallback
+ # to if the tokenizer does not have a chat template else default to tokenizer. E.g.
+ # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat
+ # template. The custom jinja template should be provided in the chat_template_jinja
+ # field.
+ chat_template: ChatTemplate | str | None
+ # Custom jinja chat template or path to jinja file. Used only if `chat_template:
+ # jinja` or empty.
+ chat_template_jinja: str | None
+ # path to source data files
+ data_files: str | list[str] | None
+ input_format: str | None
+ # name of dataset configuration to load
+ name: str | None
+ # defines the datatype when path is a file
+ ds_type: str | None
+ # For `completion` datasets only, uses the provided field instead of `text` column
+ field: str | None
+ field_human: str | None
+ field_model: str | None
+ # Key containing the messages (default: "messages")
+ field_messages: str | None
+ # Key containing the tools (default: "tools"). Must be a list[dict] and follow [JSON
+ # schema](https://json-schema.org/learn/getting-started-step-by-step).
+ field_tools: str | None
+ # Key containing the reasoning trace (default: "reasoning_content").
+ field_thinking: str | None
+ # The key the chat template expects that indicates the reasoning trace.
+ template_thinking_key: str | None
+
+ message_field_role: str | None
+
+ message_field_content: str | None
+ # Mapping of properties from the input dataset to the chat template. (default:
+ # message_property_mappings={'role':'role', 'content':'content'}) If a property exists
+ # in the template but not in this mapping, the system will attempt to load it directly
+ # from the message using the property name as the key. Example: In the mapping below,
+ # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and
+ # used as 'content' in the chat template.
+ message_property_mappings: dict[str, str] | None
+ # The key in the message turn that indicates via boolean whether tokens of a turn
+ # should be considered for training. Useful to selectively train on certain turns
+ # besides the `roles_to_train`.
+ message_field_training: str | None
+ # The key in the message turn that contains the training details. Useful to
+ # selectively train on certain tokens in a turn. The value of the key is a List[Dict]
+ # containing `begin_offset` (start character index in content), `end_offset` (end
+ # character index in content), and `train` (boolean whether to train).
+ message_field_training_detail: str | None
+ # (for Qwen3 template only) Whether to split the assistant content based on a
+ # reasoning trace inside delimited tags
+ split_thinking: bool | None
+ logprobs_field: str | None
+ temperature: float | None
+ # Roles to train on. The tokens from these roles will be considered for the loss.
+ roles_to_train: list[str] | None
+ # Which EOS tokens to train on in the conversation. Possible values are: all: train on
+ # all EOS tokens, turn (default): train on the EOS token at the end of each trainable
+ # turn, last: train on the last EOS token in the conversation
+ train_on_eos: Literal['all', 'turn', 'last'] | None
+ # Roles mapping in the messages. The format is {target_role: [source_roles]}. All
+ # source roles will be mapped to the target role. The default is: user: ["human",
+ # "user"], assistant: ["gpt", "assistant"], system: ["system"], tool: ["tool"]
+ roles: dict[str, list[str]] | None
+ # Whether to drop the system turn from the dataset. Only works with chat_template.
+ # This does not drop the default system message from chat_template if it exists. If
+ # you wish to, we recommend using a custom jinja template with the default system
+ # message removed or adding a system turn with empty content.
+ drop_system_message: bool | None
+ # Trust remote code for untrusted source
+ trust_remote_code: bool | None = False
+ # The specific revision of the dataset to use when loading from the Hugging Face Hub.
+ # This can be a commit hash, tag, or branch name. If not specified, the latest version
+ # will be used. This parameter is ignored for local datasets.
+ revision: str | None
+
+ # For DPODataset:
+ path: str | None
+ split: str | None
+ type: UserDefinedDPOType | str | None
+ # For UserDefinedDPOType:
+ field_system: str | None
+ field_prompt: str | None
+ field_chosen: str | None
+ field_rejected: str | None
+ prompt_format: str | None
+ chosen_format: str | None
+ rejected_format: str | None
+ data_files: list[str] | None
+ revision: str | None
+ field_messages: str | None
+
+ # For KTODataset:
+ path: str | None
+ split: str | None
+ type: UserDefinedKTOType | str | None
+ # For UserDefinedKTOType:
+ field_system: str | None
+ field_prompt: str | None
+ field_completion: str | None
+ field_label: bool | None
+ prompt_format: str | None
+ completion_format: str | None
+ data_files: list[str] | None
+ trust_remote_code: bool | None = False
+ revision: str | None
+
+ # For StepwiseSupervisedDataset:
+ path: str | None
+ split: str | None
+ data_files: list[str] | None
+ revision: str | None
+ step_separator: str | None
+ max_completion_length: int | None
+ train_on_last_step_only: bool | None
+
+# If false, the datasets will not be shuffled and will keep their original order in
+# `datasets`. The same applies to the `test_datasets` option and the
+# `pretraining_dataset` option. Default is true.
+shuffle_merged_datasets: bool | None = True
+# If true, each dataset in `datasets` will be shuffled before merging. This allows
+# curriculum learning strategies to be applied at the dataset level. Default is false.
+shuffle_before_merging_datasets: bool | None = False
+# Axolotl attempts to save the dataset as an arrow after packing the data together so
+# subsequent training attempts load faster, relative path
+dataset_prepared_path: str | None
+# Num shards for whole dataset
+dataset_shard_num: int | None
+# Index of shard to use for whole dataset
+dataset_shard_idx: int | None
+skip_prepare_dataset: bool | None = False
+# Number of shards to save the prepared dataset
+num_dataset_shards_to_save: int | None
+
+# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize
+pretraining_dataset: Annotated[list[PretrainingDataset | SFTDataset], MinLen(1)] | None
+ # For PretrainingDataset:
+ name: str | None
+ path: str | None
+ split: str | None = train
+ text_column: str | None = text
+ type: str | None = pretrain
+ trust_remote_code: bool | None = False
+ data_files: str | None
+ skip: int | None
+
+ # For SFTDataset:
+ # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory
+ path: str | None
+ # name of dataset split to load from
+ split: str | None
+ # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]
+ type: str | UserDefinedPrompterType | None
+ # For UserDefinedPrompterType:
+ # Custom user instruction prompt
+ system_prompt: str | None
+ # Use {system} as key to be replaced
+ system_format: str | None
+ field_system: str | None
+ field_instruction: str | None
+ field_input: str | None
+ field_output: str | None
+
+ # Customizable to be single line or multi-line. Use {instruction}/{input} as key to
+ # be replaced. 'format' can include {input}
+ format: str | None
+ # 'no_input_format' cannot include {input}
+ no_input_format: str | None
+ input_transform: str | None
+ # split dataset into N pieces (use with shards_idx)
+ shards: int | None
+ # the index of sharded dataset to use
+ shards_idx: int | None
+ # process dataset in N sequential chunks for memory efficiency (exclusive with
+ # `shards`)
+ preprocess_shards: int | None
+ conversation: str | None
+
+ # The name of the chat template to use for training, following values are supported:
+ # tokenizer_default: Uses the chat template that is available in the
+ # tokenizer_config.json. If the chat template is not available in the tokenizer, it
+ # will raise an error. This is the default.
+ # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
+ # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
+ # tokenizer_default_fallback_*: where * is the name of the chat template to fallback
+ # to if the tokenizer does not have a chat template else default to tokenizer. E.g.
+ # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat
+ # template. The custom jinja template should be provided in the chat_template_jinja
+ # field.
+ chat_template: ChatTemplate | str | None
+ # Custom jinja chat template or path to jinja file. Used only if `chat_template:
+ # jinja` or empty.
+ chat_template_jinja: str | None
+ # path to source data files
+ data_files: str | list[str] | None
+ input_format: str | None
+ # name of dataset configuration to load
+ name: str | None
+ # defines the datatype when path is a file
+ ds_type: str | None
+ # For `completion` datasets only, uses the provided field instead of `text` column
+ field: str | None
+ field_human: str | None
+ field_model: str | None
+ # Key containing the messages (default: "messages")
+ field_messages: str | None
+ # Key containing the tools (default: "tools"). Must be a list[dict] and follow [JSON
+ # schema](https://json-schema.org/learn/getting-started-step-by-step).
+ field_tools: str | None
+ # Key containing the reasoning trace (default: "reasoning_content").
+ field_thinking: str | None
+ # The key the chat template expects that indicates the reasoning trace.
+ template_thinking_key: str | None
+
+ message_field_role: str | None
+
+ message_field_content: str | None
+ # Mapping of properties from the input dataset to the chat template. (default:
+ # message_property_mappings={'role':'role', 'content':'content'}) If a property exists
+ # in the template but not in this mapping, the system will attempt to load it directly
+ # from the message using the property name as the key. Example: In the mapping below,
+ # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and
+ # used as 'content' in the chat template.
+ message_property_mappings: dict[str, str] | None
+ # The key in the message turn that indicates via boolean whether tokens of a turn
+ # should be considered for training. Useful to selectively train on certain turns
+ # besides the `roles_to_train`.
+ message_field_training: str | None
+ # The key in the message turn that contains the training details. Useful to
+ # selectively train on certain tokens in a turn. The value of the key is a List[Dict]
+ # containing `begin_offset` (start character index in content), `end_offset` (end
+ # character index in content), and `train` (boolean whether to train).
+ message_field_training_detail: str | None
+ # (for Qwen3 template only) Whether to split the assistant content based on a
+ # reasoning trace inside delimited tags
+ split_thinking: bool | None
+ logprobs_field: str | None
+ temperature: float | None
+ # Roles to train on. The tokens from these roles will be considered for the loss.
+ roles_to_train: list[str] | None
+ # Which EOS tokens to train on in the conversation. Possible values are: all: train on
+ # all EOS tokens, turn (default): train on the EOS token at the end of each trainable
+ # turn, last: train on the last EOS token in the conversation
+ train_on_eos: Literal['all', 'turn', 'last'] | None
+ # Roles mapping in the messages. The format is {target_role: [source_roles]}. All
+ # source roles will be mapped to the target role. The default is: user: ["human",
+ # "user"], assistant: ["gpt", "assistant"], system: ["system"], tool: ["tool"]
+ roles: dict[str, list[str]] | None
+ # Whether to drop the system turn from the dataset. Only works with chat_template.
+ # This does not drop the default system message from chat_template if it exists. If
+ # you wish to, we recommend using a custom jinja template with the default system
+ # message removed or adding a system turn with empty content.
+ drop_system_message: bool | None
+ # Trust remote code for untrusted source
+ trust_remote_code: bool | None = False
+ # The specific revision of the dataset to use when loading from the Hugging Face Hub.
+ # This can be a commit hash, tag, or branch name. If not specified, the latest version
+ # will be used. This parameter is ignored for local datasets.
+ revision: str | None
+
+# The maximum number of processes to use while preprocessing your input dataset. This
+# defaults to `os.cpu_count()` if not set. For Runpod VMs, it will default to number of
+# vCPUs via RUNPOD_CPU_COUNT.
+dataset_processes: int | None
+# Deduplicates datasets and test_datasets with identical entries
+dataset_exact_deduplication: bool | None
+# Keep dataset in memory while preprocessing. Only needed if cached dataset is taking
+# too much storage
+dataset_keep_in_memory: bool | None
+dataloader_pin_memory: bool | None
+dataloader_num_workers: int | None
+dataloader_prefetch_factor: int | None
+dataloader_drop_last: bool | None
+
+accelerator_config: dict[str, Any] | None
+
+remove_unused_columns: bool | None
-accelerator_config: dict[str, Any] | None
-
-remove_unused_columns: bool | None
-
-# Push prepared dataset to hub - repo_org/repo_name
-push_dataset_to_hub: str | None
-# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private
-# datasets. Required to be true when used in combination with `push_dataset_to_hub`
-hf_use_auth_token: bool | None
-
-device: Any | None
-# Passed through to transformers when loading the model when launched without
-# accelerate. Use `sequential` when training w/ model parallelism to limit memory
-device_map: Any | None
-world_size: int | None
-# Don't mess with this, it's here for accelerate and torchrun
-local_rank: int | None
-ddp: bool | None
-
-# Seed for reproducibility
-seed: int | None
-# Advanced DDP Arguments - timeout
-ddp_timeout: int | None
-# Advanced DDP Arguments - bucket cap in MB
-ddp_bucket_cap_mb: int | None
-# Advanced DDP Arguments - broadcast buffers
-ddp_broadcast_buffers: bool | None
-ddp_find_unused_parameters: bool | None
-
-# Approximate number of predictions sent to wandb depending on batch size. Enabled above
-# 0. Default is 0
-eval_table_size: int | None
-# Total number of tokens generated for predictions sent to wandb. Default is 128
-eval_max_new_tokens: int | None
-# Whether to run causal language model evaluation for metrics in
-# `eval_causal_lm_metrics`
-do_causal_lm_eval: bool | None
-# HF evaluate metrics used during evaluation. Default is ['sacrebleu', 'comet', 'ter',
-# 'chrf', 'perplexity']
-eval_causal_lm_metrics: list[str] | None
-do_bench_eval: bool | None
-bench_dataset: str | None
-bench_split: str | None
-metric_for_best_model: str | None
-greater_is_better: bool | None
-
-# High loss value, indicating the learning has broken down (a good estimate is ~2 times
-# the loss at the start of training)
-loss_watchdog_threshold: float | None
-# Number of high-loss steps in a row before the trainer aborts (default: 3)
-loss_watchdog_patience: int | None
+# Push prepared dataset to hub - repo_org/repo_name
+push_dataset_to_hub: str | None
+# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private
+# datasets. Required to be true when used in combination with `push_dataset_to_hub`
+hf_use_auth_token: bool | None
+
+device: Any | None
+# Passed through to transformers when loading the model when launched without
+# accelerate. Use `sequential` when training w/ model parallelism to limit memory
+device_map: Any | None
+world_size: int | None
+# Don't mess with this, it's here for accelerate and torchrun
+local_rank: int | None
+ddp: bool | None
+
+# Seed for reproducibility
+seed: int | None
+# Advanced DDP Arguments - timeout
+ddp_timeout: int | None
+# Advanced DDP Arguments - bucket cap in MB
+ddp_bucket_cap_mb: int | None
+# Advanced DDP Arguments - broadcast buffers
+ddp_broadcast_buffers: bool | None
+ddp_find_unused_parameters: bool | None
+
+# Approximate number of predictions sent to wandb depending on batch size. Enabled above
+# 0. Default is 0
+eval_table_size: int | None
+# Total number of tokens generated for predictions sent to wandb. Default is 128
+eval_max_new_tokens: int | None
+# Whether to run causal language model evaluation for metrics in
+# `eval_causal_lm_metrics`
+do_causal_lm_eval: bool | None
+# HF evaluate metrics used during evaluation. Default is ['sacrebleu', 'comet', 'ter',
+# 'chrf', 'perplexity']
+eval_causal_lm_metrics: list[str] | None
+do_bench_eval: bool | None
+bench_dataset: str | None
+bench_split: str | None
+metric_for_best_model: str | None
+greater_is_better: bool | None
+
+# High loss value, indicating the learning has broken down (a good estimate is ~2 times
+# the loss at the start of training)
+loss_watchdog_threshold: float | None
+# Number of high-loss steps in a row before the trainer aborts (default: 3)
+loss_watchdog_patience: int | None
+
+# Run garbage collection every `gc_steps` steps. -1 will run on epoch end and before
+# evaluations. Default is 0 (disabled).
+gc_steps: int | None
-# Run garbage collection every `gc_steps` steps. -1 will run on epoch end and before
-# evaluations. Default is 0 (disabled).
-gc_steps: int | None
-
-# Use CUDA bf16. bool or 'full' for `bf16_full_eval`, or 'auto' for automatic detection.
-# require >=ampere
-bf16: Literal['auto'] | bool | None = auto
-# Use CUDA fp16
-fp16: bool | None
-# Enable FP8 mixed precision training using TorchAO. Best used in combination with
-# torch.compile.
-fp8: bool | None
-# Enable FSDP float8 all-gather optimization for FP8 training. Can improve training
-# speed by 10-15% when FSDP is enabled.
-fp8_enable_fsdp_float8_all_gather: bool | None
-# No AMP (automatic mixed precision) - require >=ampere
-bfloat16: bool | None
-# No AMP (automatic mixed precision)
-float16: bool | None
-# Use CUDA tf32 - require >=ampere
-tf32: bool | None
-float32: bool | None
-
-# Whether to use gradient checkpointing. Available options are: true, false, 'offload',
-# 'offload_disk'.
-# https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing
-gradient_checkpointing: Literal['offload', 'offload_disk'] | bool | None = False
-# Additional kwargs to pass to the trainer for gradient checkpointing
-gradient_checkpointing_kwargs: dict[str, Any] | None
-# Whether to offload activations. Available options are: true, false, 'legacy', 'disk'.
-activation_offloading: Literal['legacy', 'disk'] | bool | None = False
-
-unfrozen_parameters: list[str] | None
-
-# The maximum length of an input to train with, this should typically be less than 2048
-# as most models have a token/context limit of 2048
-sequence_len: int = 512
-# What to do when a tokenized row exceeds sequence_len. 'drop' removes the row;
-# 'truncate' slices tensors to sequence_len. Defaults to 'drop' for backward
-# compatibility.
-excess_length_strategy: Literal['drop', 'truncate'] | None
-# The maximum length of an input for evaluation. If not specified, defaults to
-# sequence_len
-eval_sequence_len: int | None
-min_sample_len: int | None
-# maximum prompt length for RL training
-max_prompt_len: int = 512
-# Use efficient multi-packing with block diagonal attention and per sequence
-# position_ids. Recommend set to 'true'
-sample_packing: bool | None
-# The number of samples packed at a time. Increasing the following values helps with
-# packing, but usually only slightly (<%1.)
-sample_packing_group_size: int | None = 100000
-# The number of samples which can be packed into one sequence. Increase if using a large
-# sequence_len with many short samples.
-sample_packing_bin_size: int | None = 200
-# Whether to pack samples sequentially
-sample_packing_sequentially: bool | None
-# The multiprocessing start method to use for packing. Should be 'fork', 'spawn' or
-# 'forkserver'
-sample_packing_mp_start_method: str | None
-# Set to 'false' if getting errors during eval with sample_packing on
-eval_sample_packing: bool | None
-# Pad inputs so each step uses constant sized buffers. This will reduce memory
-# fragmentation and may prevent OOMs, by re-using memory more efficiently. Defaults to
-# True if `sample_packing` enabled
-pad_to_sequence_len: bool | None
-# Whether to use sequential sampling for curriculum learning
-curriculum_sampling: bool | None
-multipack_real_batches: bool | None
-
-# Use batch flattening for speedups when not using sample_packing
-batch_flattening: Literal['auto'] | bool | None
-
-use_pose: bool | None
-pose_split_on_token_ids: list[int] | None
-pose_max_context_len: int | None
-pose_num_chunks: int | None
-
-pretrain_multipack_buffer_size: int | None
-# whether to prevent cross attention for packed sequences during pretraining
-pretrain_multipack_attn: bool | None = True
-# whether to concatenate samples during pretraining
-pretraining_sample_concatenation: bool | None
-
-# Use streaming mode for loading datasets
-streaming: bool | None
-# Buffer size for multipack streaming datasets
-streaming_multipack_buffer_size: int | None = 10000
-
-# Whether to use xformers attention patch https://github.com/facebookresearch/xformers
-xformers_attention: bool | None
-# Whether to use scaled-dot-product attention https://pytorch.org/docs/stable/generated/
-# torch.nn.functional.scaled_dot_product_attention.html
-sdp_attention: bool | None
-# Shifted-sparse attention (only llama) - https://arxiv.org/pdf/2309.12307.pdf
-s2_attention: bool | None
-flex_attention: bool | None
-flex_attn_compile_kwargs: dict[str, Any] | None
-# Whether to use flash attention patch https://github.com/Dao-AILab/flash-attention
-flash_attention: bool | None
-# Whether to use flash-attention cross entropy implementation - advanced use only
-flash_attn_cross_entropy: bool | None
-# Whether to use flash-attention rms norm implementation - advanced use only
-flash_attn_rms_norm: bool | None
-# Whether to fuse part of the MLP into a single operation
-flash_attn_fuse_mlp: bool | None
-# Whether to use bettertransformers
-flash_optimum: bool | None
-
-eager_attention: bool | None
-
-# Specify a custom attention implementation, used mostly for kernels.
-attn_implementation: str | None
-
-unsloth_cross_entropy_loss: bool | None
-unsloth_lora_mlp: bool | None
-unsloth_lora_qkv: bool | None
-unsloth_lora_o: bool | None
-unsloth_rms_norm: bool | None
-unsloth_rope: bool | None
-
-# Apply custom LoRA autograd functions and activation function Triton kernels for speed
-# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html
-lora_mlp_kernel: bool | None
-# Apply custom LoRA autograd functions and activation function Triton kernels for speed
-# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html
-lora_qkv_kernel: bool | None
-# Apply custom LoRA autograd functions and activation function Triton kernels for speed
-# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html
-lora_o_kernel: bool | None
-
-# Whether to use chunked cross entropy loss for memory efficiency
-chunked_cross_entropy: bool | None
-# Number of chunks to use for chunked cross entropy loss
-chunked_cross_entropy_num_chunks: int | None
-
-# Whether to use ALST tiled mlp for memory efficient long context
-tiled_mlp: bool | None
+# Use CUDA bf16. bool or 'full' for `bf16_full_eval`, or 'auto' for automatic detection.
+# require >=ampere
+bf16: Literal['auto'] | bool | None = auto
+# Use CUDA fp16
+fp16: bool | None
+# Enable FP8 mixed precision training using TorchAO. Best used in combination with
+# torch.compile.
+fp8: bool | None
+# Enable FSDP float8 all-gather optimization for FP8 training. Can improve training
+# speed by 10-15% when FSDP is enabled.
+fp8_enable_fsdp_float8_all_gather: bool | None
+# No AMP (automatic mixed precision) - require >=ampere
+bfloat16: bool | None
+# No AMP (automatic mixed precision)
+float16: bool | None
+# Use CUDA tf32 - require >=ampere
+tf32: bool | None
+float32: bool | None
+
+# Whether to use gradient checkpointing. Available options are: true, false, 'offload',
+# 'offload_disk'.
+# https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing
+gradient_checkpointing: Literal['offload', 'offload_disk'] | bool | None = False
+# Additional kwargs to pass to the trainer for gradient checkpointing
+gradient_checkpointing_kwargs: dict[str, Any] | None
+# Whether to offload activations. Available options are: true, false, 'legacy', 'disk'.
+activation_offloading: Literal['legacy', 'disk'] | bool | None = False
+
+unfrozen_parameters: list[str] | None
+
+# The maximum length of an input to train with, this should typically be less than 2048
+# as most models have a token/context limit of 2048
+sequence_len: int = 512
+# What to do when a tokenized row exceeds sequence_len. 'drop' removes the row;
+# 'truncate' slices tensors to sequence_len. Defaults to 'drop' for backward
+# compatibility.
+excess_length_strategy: Literal['drop', 'truncate'] | None
+# The maximum length of an input for evaluation. If not specified, defaults to
+# sequence_len
+eval_sequence_len: int | None
+min_sample_len: int | None
+# maximum prompt length for RL training
+max_prompt_len: int = 512
+# Use efficient multi-packing with block diagonal attention and per sequence
+# position_ids. Recommend set to 'true'
+sample_packing: bool | None
+# The number of samples packed at a time. Increasing the following values helps with
+# packing, but usually only slightly (<%1.)
+sample_packing_group_size: int | None = 100000
+# The number of samples which can be packed into one sequence. Increase if using a large
+# sequence_len with many short samples.
+sample_packing_bin_size: int | None = 200
+# Whether to pack samples sequentially
+sample_packing_sequentially: bool | None
+# The multiprocessing start method to use for packing. Should be 'fork', 'spawn' or
+# 'forkserver'
+sample_packing_mp_start_method: str | None
+# Set to 'false' if getting errors during eval with sample_packing on
+eval_sample_packing: bool | None
+# Pad inputs so each step uses constant sized buffers. This will reduce memory
+# fragmentation and may prevent OOMs, by re-using memory more efficiently. Defaults to
+# True if `sample_packing` enabled
+pad_to_sequence_len: bool | None
+# Whether to use sequential sampling for curriculum learning
+curriculum_sampling: bool | None
+multipack_real_batches: bool | None
+
+# Use batch flattening for speedups when not using sample_packing
+batch_flattening: Literal['auto'] | bool | None
+
+use_pose: bool | None
+pose_split_on_token_ids: list[int] | None
+pose_max_context_len: int | None
+pose_num_chunks: int | None
+
+pretrain_multipack_buffer_size: int | None
+# whether to prevent cross attention for packed sequences during pretraining
+pretrain_multipack_attn: bool | None = True
+# whether to concatenate samples during pretraining
+pretraining_sample_concatenation: bool | None
+
+# Use streaming mode for loading datasets
+streaming: bool | None
+# Buffer size for multipack streaming datasets
+streaming_multipack_buffer_size: int | None = 10000
+
+# Whether to use xformers attention patch https://github.com/facebookresearch/xformers
+xformers_attention: bool | None
+# Whether to use scaled-dot-product attention https://pytorch.org/docs/stable/generated/
+# torch.nn.functional.scaled_dot_product_attention.html
+sdp_attention: bool | None
+# Shifted-sparse attention (only llama) - https://arxiv.org/pdf/2309.12307.pdf
+s2_attention: bool | None
+flex_attention: bool | None
+flex_attn_compile_kwargs: dict[str, Any] | None
+# Whether to use flash attention patch https://github.com/Dao-AILab/flash-attention
+flash_attention: bool | None
+# Whether to use flash-attention cross entropy implementation - advanced use only
+flash_attn_cross_entropy: bool | None
+# Whether to use flash-attention rms norm implementation - advanced use only
+flash_attn_rms_norm: bool | None
+# Whether to fuse part of the MLP into a single operation
+flash_attn_fuse_mlp: bool | None
+# Whether to use bettertransformers
+flash_optimum: bool | None
+
+eager_attention: bool | None
+
+# Specify a custom attention implementation, used mostly for kernels.
+attn_implementation: str | None
+
+unsloth_cross_entropy_loss: bool | None
+unsloth_lora_mlp: bool | None
+unsloth_lora_qkv: bool | None
+unsloth_lora_o: bool | None
+unsloth_rms_norm: bool | None
+unsloth_rope: bool | None
+
+# Apply custom LoRA autograd functions and activation function Triton kernels for speed
+# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html
+lora_mlp_kernel: bool | None
+# Apply custom LoRA autograd functions and activation function Triton kernels for speed
+# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html
+lora_qkv_kernel: bool | None
+# Apply custom LoRA autograd functions and activation function Triton kernels for speed
+# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html
+lora_o_kernel: bool | None
+
+# Whether to use chunked cross entropy loss for memory efficiency
+chunked_cross_entropy: bool | None
+# Number of chunks to use for chunked cross entropy loss
+chunked_cross_entropy_num_chunks: int | None
+
+# Whether to use ALST tiled mlp for memory efficient long context
+tiled_mlp: bool | None
+
+# Number of shards to use for ALST tiled mlp. If unset, it will be set based on
+# seqlen/hidden_size
+tiled_mlp_num_shards: int | None
-# Number of shards to use for ALST tiled mlp. If unset, it will be set based on
-# seqlen/hidden_size
-tiled_mlp_num_shards: int | None
+# Whether to use original mlp for ALST tiled mlp. Otherwise uses a generic MLP based on
+# llama.
+tiled_mlp_use_original_mlp: bool | None = True
-# Whether to use original mlp for ALST tiled mlp. Otherwise uses a generic MLP based on
-# llama.
-tiled_mlp_use_original_mlp: bool | None = True
-
-llama4_linearized_experts: bool | None
-
-# Deepspeed config path. e.g., deepspeed_configs/zero3.json
-deepspeed: str | dict[str, Any] | None
-# Whether to use deepcompile for faster training with deepspeed
-deepcompile: bool | None
-# FSDP configuration
-fsdp: list[str] | None
-
-# FSDP configuration options
-fsdp_config: dict[str, Any] | None
-# FSDP version
-fsdp_version: int | None
-fsdp_final_state_dict_type: Literal['FULL_STATE_DICT', 'LOCAL_STATE_DICT', 'SHARDED_STATE_DICT'] | None
+llama4_linearized_experts: bool | None
+
+# Deepspeed config path. e.g., deepspeed_configs/zero3.json
+deepspeed: str | dict[str, Any] | None
+# Whether to use deepcompile for faster training with deepspeed
+deepcompile: bool | None
+# FSDP configuration
+fsdp: list[str] | None
+
+# FSDP configuration options
+fsdp_config: dict[str, Any] | None
+# FSDP version
+fsdp_version: int | None
+fsdp_final_state_dict_type: Literal['FULL_STATE_DICT', 'LOCAL_STATE_DICT', 'SHARDED_STATE_DICT'] | None
+
+# How much of the dataset to set aside as evaluation. 1 = 100%, 0.50 = 50%, etc. 0 for
+# no eval.
+val_set_size: float | None = 0.0
-# How much of the dataset to set aside as evaluation. 1 = 100%, 0.50 = 50%, etc. 0 for
-# no eval.
-val_set_size: float | None = 0.0
-
-# Number of devices to shard across. If not set, will use all available devices.
-dp_shard_size: int | None
-# Number of devices to replicate across.
-dp_replicate_size: int | None
-# Deprecated: use `context_parallel_size` instead
-sequence_parallel_degree: int | None
-# Set to a divisor of the number of GPUs available to split sequences into chunks of
-# equal size. Use in long context training to prevent OOM when sequences cannot fit into
-# a single GPU's VRAM. E.g., if 4 GPUs are available, set this value to 2 to split each
-# sequence into two equal-sized subsequences, or set to 4 to split into four equal-sized
-# subsequences. See https://docs.axolotl.ai/docs/sequence_parallelism.html for more
-# details.
-context_parallel_size: int | None
-# Optional; strides across the key dimension. Larger values use more memory but should
-# make training faster. Must evenly divide the number of KV heads in your model.
-heads_k_stride: int | None
-# One of 'varlen_llama3', 'batch_ring', 'batch_zigzag', 'batch_stripe'. Defaults to
-# 'varlen_llama3' in the sample packing case, and 'batch_ring' in the non-sample packing
-# case.
-ring_attn_func: RingAttnFunc | None
-# Number of tensor parallel processes in TP group. Only supported with DeepSpeed AutoTP.
-tensor_parallel_size: int | None
-
-# Add or change special tokens. If you add tokens here, you don't need to add them to
-# the `tokens` list.
-special_tokens: SpecialTokensConfig | None
- # For SpecialTokensConfig:
- bos_token: str | None
- eos_token: str | None
- pad_token: str | None
- unk_token: str | None
- additional_special_tokens: list[str] | None
-
-# Add extra tokens to the tokenizer
-tokens: list[str] | None
-# Mapping token_id to new_token_string to override reserved added_tokens in the
-# tokenizer. Only works for tokens that are not part of the base vocab (aka are
-# added_tokens). Can be checked if they exist in tokenizer.json added_tokens.
-added_tokens_overrides: dict[int, str] | None
-
-# Whether to use torch.compile and which backend to use. setting to `auto` will enable
-# torch compile when torch>=2.6.0
-torch_compile: Literal['auto'] | bool | None
-# Backend to use for torch.compile
-torch_compile_backend: str | None
-torch_compile_mode: Literal['default', 'reduce-overhead', 'max-autotune'] | None
-
-# Maximum number of iterations to train for. It precedes num_epochs which means that if
-# both are set, num_epochs will not be guaranteed. e.g., when 1 epoch is 1000 steps =>
-# `num_epochs: 2` and `max_steps: 100` will train for 100 steps
-max_steps: int | None
-# Number of warmup steps. Cannot use with warmup_ratio
-warmup_steps: int | None
-# Warmup ratio. Cannot use with warmup_steps
-warmup_ratio: float | None
-# Leave empty to eval at each epoch, integer for every N steps. float for fraction of
-# total steps
-eval_steps: int | float | None
-# Number of times per epoch to run evals, mutually exclusive with eval_steps
-evals_per_epoch: int | None
-# Set to `no` to skip evaluation, `epoch` at end of each epoch, leave empty to infer
-# from `eval_steps`
-eval_strategy: str | None
-
-# Leave empty to save at each epoch, integer for every N steps. float for fraction of
-# total steps
-save_steps: int | float | None
-# Number of times per epoch to save a checkpoint, mutually exclusive with save_steps
-saves_per_epoch: int | None
-# Set to `no` to skip checkpoint saves, `epoch` at end of each epoch, `best` when better
-# result is achieved, leave empty to infer from `save_steps`
-save_strategy: str | None
-# Checkpoints saved at a time
-save_total_limit: int | None
-# Whether to checkpoint a model after the first step of training. Defaults to False.
-save_first_step: bool | None
-
-# Logging frequency
-logging_steps: int | None
-# Stop training after this many evaluation losses have increased in a row. https://huggi
-# ngface.co/transformers/v4.2.2/_modules/transformers/trainer_callback.html#EarlyStoppin
-# gCallback
-early_stopping_patience: int | None
-load_best_model_at_end: bool | None = False
-# Save only the model weights, skipping the optimizer. Using this means you can't resume
-# from checkpoints.
-save_only_model: bool | None = False
-# Use tensorboard for logging
-use_tensorboard: bool | None
-# Enable the pytorch profiler to capture the first N steps of training to the
-# output_dir. see https://pytorch.org/blog/understanding-gpu-memory-1/ for more
-# information. Snapshots can be visualized @ https://pytorch.org/memory_viz
-profiler_steps: int | None
-# Which step to start the profiler at. Useful for only capturing a few steps mid-run.
-profiler_steps_start: int | None = 0
-# bool of whether to report tokens per second at the end of training. This is not
-# supported with pre-training datasets.
-include_tokens_per_second: bool | None
-# bool of whether to report tokens per second per-gpu during training by measuring
-# throughput of non-padding tokens.
-include_tkps: bool | None = True
-# NEFT https://arxiv.org/abs/2310.05914, set this to a number (paper default is 5) to
-# add noise to embeddings. Currently only supported on Llama and Mistral
-neftune_noise_alpha: float | None
-
-# Parameter controlling the relative ratio loss weight in the ORPO loss. Passed to
-# `beta` in `ORPOConfig` due to trl mapping.
-orpo_alpha: float | None
-# Weighting of NLL term in loss from RPO paper
-rpo_alpha: float | None
-# Target reward margin for the SimPO loss
-simpo_gamma: float | None
-# Weight of the BC regularizer
-cpo_alpha: float | None
-
-# Factor for desirable loss term in KTO loss
-kto_desirable_weight: float | None
-# Factor for undesirable loss term in KTO loss
-kto_undesirable_weight: float | None
-# The beta parameter for the RL training
-rl_beta: float | None
-
-# Defines the max memory usage per gpu on the system. Passed through to transformers
-# when loading the model.
-max_memory: dict[int | Literal['cpu', 'disk'], int | str] | None
-# Limit the memory for all available GPUs to this amount (if an integer, expressed in
-# gigabytes); default: unset
-gpu_memory_limit: int | str | None
-# Whether to use low_cpu_mem_usage
-low_cpu_mem_usage: bool | None
-
-# The name of the chat template to use for training, following values are supported:
-# tokenizer_default: Uses the chat template that is available in the
-# tokenizer_config.json. If the chat template is not available in the tokenizer, it will
-# raise an error. This is the default value.
-# alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
-# are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
-# tokenizer_default_fallback_*: where * is the name of the chat template to fallback to.
-# E.g. tokenizer_default_fallback_chatml. This is useful when the chat template is not
-# available in the tokenizer. jinja: Uses a custom jinja template for the chat template.
-# The custom jinja template should be provided in the chat_template_jinja field. The
-# selected chat template will be saved to the tokenizer_config.json for easier
-# inferencing
-chat_template: ChatTemplate | Annotated[str, StringConstraints(pattern='^tokenizer_default_fallback_')] | None
-# Custom jinja template or path to jinja file for chat template. This will be only used
-# if chat_template is set to `jinja` or `null` (in which case chat_template is
-# automatically set to `jinja`). Default is null.
-chat_template_jinja: str | None
-# Additional kwargs to pass to the chat template. This is useful for customizing the
-# chat template. For example, you can pass `thinking=False` to add a generation prompt
-# to the chat template.
-chat_template_kwargs: dict[str, Any] | None
-# Custom EOT (End-of-Turn) tokens to mask/unmask during training. These tokens mark the
-# boundaries between conversation turns. For example: ['/INST', '</s>',
-# '[/SYSTEM_PROMPT]']. If not specified, defaults to just the model's eos_token. This is
-# useful for templates that use multiple delimiter tokens.
-eot_tokens: list[str] | None
-# Changes the default system message. Currently only supports chatml.
-default_system_message: str | None
+# Number of devices to shard across. If not set, will use all available devices.
+dp_shard_size: int | None
+# Number of devices to replicate across.
+dp_replicate_size: int | None
+# Deprecated: use `context_parallel_size` instead
+sequence_parallel_degree: int | None
+# Set to a divisor of the number of GPUs available to split sequences into chunks of
+# equal size. Use in long context training to prevent OOM when sequences cannot fit into
+# a single GPU's VRAM. E.g., if 4 GPUs are available, set this value to 2 to split each
+# sequence into two equal-sized subsequences, or set to 4 to split into four equal-sized
+# subsequences. See https://docs.axolotl.ai/docs/sequence_parallelism.html for more
+# details.
+context_parallel_size: int | None
+# Optional; strides across the key dimension. Larger values use more memory but should
+# make training faster. Must evenly divide the number of KV heads in your model.
+heads_k_stride: int | None
+# One of 'varlen_llama3', 'batch_ring', 'batch_zigzag', 'batch_stripe'. Defaults to
+# 'varlen_llama3' in the sample packing case, and 'batch_ring' in the non-sample packing
+# case.
+ring_attn_func: RingAttnFunc | None
+# Number of tensor parallel processes in TP group. Only supported with DeepSpeed AutoTP.
+tensor_parallel_size: int | None
+
+# Add or change special tokens. If you add tokens here, you don't need to add them to
+# the `tokens` list.
+special_tokens: SpecialTokensConfig | None
+ # For SpecialTokensConfig:
+ bos_token: str | None
+ eos_token: str | None
+ pad_token: str | None
+ unk_token: str | None
+ additional_special_tokens: list[str] | None
+
+# Add extra tokens to the tokenizer
+tokens: list[str] | None
+# Mapping token_id to new_token_string to override reserved added_tokens in the
+# tokenizer. Only works for tokens that are not part of the base vocab (aka are
+# added_tokens). Can be checked if they exist in tokenizer.json added_tokens.
+added_tokens_overrides: dict[int, str] | None
+
+# Whether to use torch.compile and which backend to use. setting to `auto` will enable
+# torch compile when torch>=2.6.0
+torch_compile: Literal['auto'] | bool | None
+# Backend to use for torch.compile
+torch_compile_backend: str | None
+torch_compile_mode: Literal['default', 'reduce-overhead', 'max-autotune'] | None
+
+# Maximum number of iterations to train for. It precedes num_epochs which means that if
+# both are set, num_epochs will not be guaranteed. e.g., when 1 epoch is 1000 steps =>
+# `num_epochs: 2` and `max_steps: 100` will train for 100 steps
+max_steps: int | None
+# Number of warmup steps. Cannot use with warmup_ratio
+warmup_steps: int | None
+# Warmup ratio. Cannot use with warmup_steps
+warmup_ratio: float | None
+# Leave empty to eval at each epoch, integer for every N steps. float for fraction of
+# total steps
+eval_steps: int | float | None
+# Number of times per epoch to run evals, mutually exclusive with eval_steps
+evals_per_epoch: int | None
+# Set to `no` to skip evaluation, `epoch` at end of each epoch, leave empty to infer
+# from `eval_steps`
+eval_strategy: str | None
+
+# Leave empty to save at each epoch, integer for every N steps. float for fraction of
+# total steps
+save_steps: int | float | None
+# Number of times per epoch to save a checkpoint, mutually exclusive with save_steps
+saves_per_epoch: int | None
+# Set to `no` to skip checkpoint saves, `epoch` at end of each epoch, `best` when better
+# result is achieved, leave empty to infer from `save_steps`
+save_strategy: str | None
+# Checkpoints saved at a time
+save_total_limit: int | None
+# Whether to checkpoint a model after the first step of training. Defaults to False.
+save_first_step: bool | None
+
+# Logging frequency
+logging_steps: int | None
+# Stop training after this many evaluation losses have increased in a row. https://huggi
+# ngface.co/transformers/v4.2.2/_modules/transformers/trainer_callback.html#EarlyStoppin
+# gCallback
+early_stopping_patience: int | None
+load_best_model_at_end: bool | None = False
+# Save only the model weights, skipping the optimizer. Using this means you can't resume
+# from checkpoints.
+save_only_model: bool | None = False
+# Use tensorboard for logging
+use_tensorboard: bool | None
+# Enable the pytorch profiler to capture the first N steps of training to the
+# output_dir. see https://pytorch.org/blog/understanding-gpu-memory-1/ for more
+# information. Snapshots can be visualized @ https://pytorch.org/memory_viz
+profiler_steps: int | None
+# Which step to start the profiler at. Useful for only capturing a few steps mid-run.
+profiler_steps_start: int | None = 0
+# bool of whether to report tokens per second at the end of training. This is not
+# supported with pre-training datasets.
+include_tokens_per_second: bool | None
+# bool of whether to report tokens per second per-gpu during training by measuring
+# throughput of non-padding tokens.
+include_tkps: bool | None = True
+# NEFT https://arxiv.org/abs/2310.05914, set this to a number (paper default is 5) to
+# add noise to embeddings. Currently only supported on Llama and Mistral
+neftune_noise_alpha: float | None
+
+# Parameter controlling the relative ratio loss weight in the ORPO loss. Passed to
+# `beta` in `ORPOConfig` due to trl mapping.
+orpo_alpha: float | None
+# Weighting of NLL term in loss from RPO paper
+rpo_alpha: float | None
+# Target reward margin for the SimPO loss
+simpo_gamma: float | None
+# Weight of the BC regularizer
+cpo_alpha: float | None
+
+# Factor for desirable loss term in KTO loss
+kto_desirable_weight: float | None
+# Factor for undesirable loss term in KTO loss
+kto_undesirable_weight: float | None
+# The beta parameter for the RL training
+rl_beta: float | None
+
+# Defines the max memory usage per gpu on the system. Passed through to transformers
+# when loading the model.
+max_memory: dict[int | Literal['cpu', 'disk'], int | str] | None
+# Limit the memory for all available GPUs to this amount (if an integer, expressed in
+# gigabytes); default: unset
+gpu_memory_limit: int | str | None
+# Whether to use low_cpu_mem_usage
+low_cpu_mem_usage: bool | None
+
+# The name of the chat template to use for training, following values are supported:
+# tokenizer_default: Uses the chat template that is available in the
+# tokenizer_config.json. If the chat template is not available in the tokenizer, it will
+# raise an error. This is the default value.
+# alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates
+# are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.
+# tokenizer_default_fallback_*: where * is the name of the chat template to fallback to.
+# E.g. tokenizer_default_fallback_chatml. This is useful when the chat template is not
+# available in the tokenizer. jinja: Uses a custom jinja template for the chat template.
+# The custom jinja template should be provided in the chat_template_jinja field. The
+# selected chat template will be saved to the tokenizer_config.json for easier
+# inferencing
+chat_template: ChatTemplate | Annotated[str, StringConstraints(pattern='^tokenizer_default_fallback_')] | None
+# Custom jinja template or path to jinja file for chat template. This will be only used
+# if chat_template is set to `jinja` or `null` (in which case chat_template is
+# automatically set to `jinja`). Default is null.
+chat_template_jinja: str | None
+# Additional kwargs to pass to the chat template. This is useful for customizing the
+# chat template. For example, you can pass `thinking=False` to add a generation prompt
+# to the chat template.
+chat_template_kwargs: dict[str, Any] | None
+# Custom EOT (End-of-Turn) tokens to mask/unmask during training. These tokens mark the
+# boundaries between conversation turns. For example: ['/INST', '</s>',
+# '[/SYSTEM_PROMPT]']. If not specified, defaults to just the model's eos_token. This is
+# useful for templates that use multiple delimiter tokens.
+eot_tokens: list[str] | None
+# Changes the default system message. Currently only supports chatml.
+default_system_message: str | None
+
+# Token index or indices to adjust embedding weights to the mean of the other tokens.
+# This is useful when the model has untrained embeddings.
+fix_untrained_tokens: int | list[int] | None
-# Token index or indices to adjust embedding weights to the mean of the other tokens.
-# This is useful when the model has untrained embeddings.
-fix_untrained_tokens: int | list[int] | None
-
-is_preprocess: bool | None
-preprocess_iterable: bool | None
-
-# Total number of tokens - internal use
-total_num_tokens: int | None
-total_supervised_tokens: int | None
-# You can set these packing optimizations AFTER starting a training at least once. The
-# trainer will provide recommended values for these values.
-sample_packing_eff_est: float | None
-axolotl_config_path: str | None
-
-# Internal use only - Used to identify which the model is based on
-is_falcon_derived_model: bool | None
-# Internal use only - Used to identify which the model is based on
-is_llama_derived_model: bool | None
-# Internal use only - Used to identify which the model is based on. Please note that if
-# you set this to true, `padding_side` will be set to 'left' by default
-is_mistral_derived_model: bool | None
-# Internal use only - Used to identify which the model is based on
-is_qwen_derived_model: bool | None
-
-# Add plugins to extend the pipeline. See `src/axolotl/integrations` for the available
-# plugins or doc below for more details.
-# https://docs.axolotl.ai/docs/custom_integrations.html
-plugins: list[str] | None
-
-# This is the huggingface model that contains *.pt, *.safetensors, or *.bin files. This
-# can also be a relative path to a model on disk
-base_model: str (required)
-# If the base_model repo on hf hub doesn't include configuration .json files, You can
-# set that here, or leave this empty to default to base_model
-base_model_config: str | None
-cls_model_config: str | None
-# Optional tokenizer configuration path in case you want to use a different tokenizer
-# than the one defined in the base model
-tokenizer_config: str | None
-# use_fast option for tokenizer loading from_pretrained, default to True
-tokenizer_use_fast: bool | None
-# Whether to use the legacy tokenizer setting, defaults to True
-tokenizer_legacy: bool | None
-# Whether to use mistral-common tokenizer. If set to True, it will use the mistral-
-# common tokenizer.
-tokenizer_use_mistral_common: bool | None
-# Corresponding tokenizer for the model AutoTokenizer is a good choice
-tokenizer_type: str | None
-# transformers processor class
-processor_type: str | None
-# Whether to save jinja files for tokenizer, transformers default is True
-tokenizer_save_jinja_files: bool | None = True
-# Trust remote code for untrusted source
-trust_remote_code: bool | None
+is_preprocess: bool | None
+preprocess_iterable: bool | None
+
+# Total number of tokens - internal use
+total_num_tokens: int | None
+total_supervised_tokens: int | None
+# You can set these packing optimizations AFTER starting a training at least once. The
+# trainer will provide recommended values for these values.
+sample_packing_eff_est: float | None
+axolotl_config_path: str | None
+
+# Internal use only - Used to identify which the model is based on
+is_falcon_derived_model: bool | None
+# Internal use only - Used to identify which the model is based on
+is_llama_derived_model: bool | None
+# Internal use only - Used to identify which the model is based on. Please note that if
+# you set this to true, `padding_side` will be set to 'left' by default
+is_mistral_derived_model: bool | None
+# Internal use only - Used to identify which the model is based on
+is_qwen_derived_model: bool | None
+
+# Add plugins to extend the pipeline. See `src/axolotl/integrations` for the available
+# plugins or doc below for more details.
+# https://docs.axolotl.ai/docs/custom_integrations.html
+plugins: list[str] | None
+
+# This is the huggingface model that contains *.pt, *.safetensors, or *.bin files. This
+# can also be a relative path to a model on disk
+base_model: str (required)
+# If the base_model repo on hf hub doesn't include configuration .json files, You can
+# set that here, or leave this empty to default to base_model
+base_model_config: str | None
+cls_model_config: str | None
+# Optional tokenizer configuration path in case you want to use a different tokenizer
+# than the one defined in the base model
+tokenizer_config: str | None
+# use_fast option for tokenizer loading from_pretrained, default to True
+tokenizer_use_fast: bool | None
+# Whether to use the legacy tokenizer setting, defaults to True
+tokenizer_legacy: bool | None
+# Whether to use mistral-common tokenizer. If set to True, it will use the mistral-
+# common tokenizer.
+tokenizer_use_mistral_common: bool | None
+# Corresponding tokenizer for the model AutoTokenizer is a good choice
+tokenizer_type: str | None
+# transformers processor class
+processor_type: str | None
+# Whether to save jinja files for tokenizer, transformers default is True
+tokenizer_save_jinja_files: bool | None = True
+# Trust remote code for untrusted source
+trust_remote_code: bool | None
+
+# Don't move the model to the device before sharding. Set to `false` to revert to legacy
+# behavior.
+experimental_skip_move_to_device: bool | None = True
-# Don't move the model to the device before sharding. Set to `false` to revert to legacy
-# behavior.
-experimental_skip_move_to_device: bool | None = True
-
-# Use custom kernels, e.g. MegaBlocks.
-use_kernels: bool | None
-
-# Model loading quantization config
-model_quantization_config: Literal['Mxfp4Config'] | None
-# kwargs for model quantization config
-model_quantization_config_kwargs: dict[str, Any] | None
-
-# Where to save the full-finetuned model to
-output_dir: str = ./model-out
-# push checkpoints to hub
-hub_model_id: str | None
-# how to push checkpoints to hub
-hub_strategy: str | None
-# Save model as safetensors (require safetensors package). Default True
-save_safetensors: bool | None = True
-
-# This will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer
-load_in_8bit: bool | None = False
-# Use bitsandbytes 4 bit
-load_in_4bit: bool | None = False
-
-# If you want to use 'lora' or 'qlora' or leave blank to train all parameters in
-# original model
-adapter: str | None
-# If you already have a lora model trained that you want to load, put that here. This
-# means after training, if you want to test the model, you should set this to the value
-# of `output_dir`. Note that if you merge an adapter to the base model, a new
-# subdirectory `merged` will be created under the `output_dir`.
-lora_model_dir: str | None
-lora_r: int | None
-lora_alpha: int | None
-lora_fan_in_fan_out: bool | None
-lora_target_modules: str | list[str] | None
-lora_target_parameters: str | list[str] | None
-# If true, will target all linear modules
-lora_target_linear: bool | None
-# If you added new tokens to the tokenizer, you may need to save some LoRA modules
-# because they need to know the new tokens. For LLaMA and Mistral, you need to save
-# `embed_tokens` and `lm_head`. It may vary for other models. `embed_tokens` converts
-# tokens to embeddings, and `lm_head` converts embeddings to token probabilities.
-lora_modules_to_save: list[str] | None
-lora_dropout: float | None = 0.0
-# The layer indices to transform, otherwise, apply to all layers
-peft_layers_to_transform: list[int] | None
-peft_layers_pattern: list[str] | None
-
-peft: PeftConfig | None
- # For PeftConfig:
- # Configuration options for loftq initialization for LoRA
- loftq_config: LoftQConfig | None
- # For LoftQConfig:
- # typically 4 bits
- loftq_bits: int = 4
-
-# Whether to use DoRA.
-peft_use_dora: bool | None
-# Whether to use RSLoRA.
-peft_use_rslora: bool | None
-# List of layer indices to replicate.
-peft_layer_replication: list[tuple[int, int]] | None
-# How to initialize LoRA weights. Default to True which is MS original implementation.
-peft_init_lora_weights: bool | str | None
-# A list of token indices to fine-tune on the `embed_tokens` layer. Otherwise, a dict
-# mapping an embedding layer name to its trainable token indices. See
-# https://huggingface.co/docs/peft/v0.17.0/en/developer_guides/lora#efficiently-train-
-# tokens-alongside-lora
-peft_trainable_token_indices: list[int] | dict[str, list[int]] | None
-
-# load qlora model in sharded format for FSDP using answer.ai technique.
-qlora_sharded_model_loading: bool | None = False
-# Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it
-# takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge
-lora_on_cpu: bool | None
-# Whether you are training a 4-bit GPTQ quantized model
-gptq: bool | None
-# optional overrides to the bnb 4bit quantization configuration
-bnb_config_kwargs: dict[str, Any] | None
-
-# loraplus learning rate ratio lr_B / lr_A. Recommended value is 2^4.
-loraplus_lr_ratio: float | None
-# loraplus learning rate for lora embedding layers. Default value is 1e-6.
-loraplus_lr_embedding: float | None = 1e-06
-
-merge_lora: bool | None
-
-# Whether to use ReLoRA. Use with jagged_restart_*steps options.
-relora: bool | None
-# threshold for optimizer magnitude when pruning
-relora_prune_ratio: float | None
-# True to perform lora weight merges on cpu during restarts, for modest gpu memory
-# savings
-relora_cpu_offload: bool | None
-
-# how often to reset for jagged restarts
-jagged_restart_steps: int | None
-# how many warmup steps to take after reset for jagged restarts
-jagged_restart_warmup_steps: int | None
-# how many anneal steps to take before reset for jagged restarts
-jagged_restart_anneal_steps: int | None
-
-# If greater than 1, backpropagation will be skipped and the gradients will be
-# accumulated for the given number of steps.
-gradient_accumulation_steps: int | None = 1
-# The number of samples to include in each batch. This is the number of samples sent to
-# each GPU. Batch size per gpu = micro_batch_size * gradient_accumulation_steps
-micro_batch_size: int | None = 1
-# Total batch size, we do not recommended setting this manually
-batch_size: int | None
-# per gpu micro batch size for evals, defaults to value of micro_batch_size
-eval_batch_size: int | None
+# Use custom kernels, e.g. MegaBlocks.
+use_kernels: bool | None
+
+# Model loading quantization config
+model_quantization_config: Literal['Mxfp4Config'] | None
+# kwargs for model quantization config
+model_quantization_config_kwargs: dict[str, Any] | None
+
+# Where to save the full-finetuned model to
+output_dir: str = ./model-out
+# push checkpoints to hub
+hub_model_id: str | None
+# how to push checkpoints to hub
+hub_strategy: str | None
+# Save model as safetensors (require safetensors package). Default True
+save_safetensors: bool | None = True
+
+# This will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer
+load_in_8bit: bool | None = False
+# Use bitsandbytes 4 bit
+load_in_4bit: bool | None = False
+
+# If you want to use 'lora' or 'qlora' or leave blank to train all parameters in
+# original model
+adapter: str | None
+# If you already have a lora model trained that you want to load, put that here. This
+# means after training, if you want to test the model, you should set this to the value
+# of `output_dir`. Note that if you merge an adapter to the base model, a new
+# subdirectory `merged` will be created under the `output_dir`.
+lora_model_dir: str | None
+lora_r: int | None
+lora_alpha: int | None
+lora_fan_in_fan_out: bool | None
+lora_target_modules: str | list[str] | None
+lora_target_parameters: str | list[str] | None
+# If true, will target all linear modules
+lora_target_linear: bool | None
+# If you added new tokens to the tokenizer, you may need to save some LoRA modules
+# because they need to know the new tokens. For LLaMA and Mistral, you need to save
+# `embed_tokens` and `lm_head`. It may vary for other models. `embed_tokens` converts
+# tokens to embeddings, and `lm_head` converts embeddings to token probabilities.
+lora_modules_to_save: list[str] | None
+lora_dropout: float | None = 0.0
+# The layer indices to transform, otherwise, apply to all layers
+peft_layers_to_transform: list[int] | None
+peft_layers_pattern: list[str] | None
+
+peft: PeftConfig | None
+ # For PeftConfig:
+ # Configuration options for loftq initialization for LoRA
+ loftq_config: LoftQConfig | None
+ # For LoftQConfig:
+ # typically 4 bits
+ loftq_bits: int = 4
+
+# Whether to use DoRA.
+peft_use_dora: bool | None
+# Whether to use RSLoRA.
+peft_use_rslora: bool | None
+# List of layer indices to replicate.
+peft_layer_replication: list[tuple[int, int]] | None
+# How to initialize LoRA weights. Default to True which is MS original implementation.
+peft_init_lora_weights: bool | str | None
+# A list of token indices to fine-tune on the `embed_tokens` layer. Otherwise, a dict
+# mapping an embedding layer name to its trainable token indices. See
+# https://huggingface.co/docs/peft/v0.17.0/en/developer_guides/lora#efficiently-train-
+# tokens-alongside-lora
+peft_trainable_token_indices: list[int] | dict[str, list[int]] | None
+
+# load qlora model in sharded format for FSDP using answer.ai technique.
+qlora_sharded_model_loading: bool | None = False
+# Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it
+# takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge
+lora_on_cpu: bool | None
+# Whether you are training a 4-bit GPTQ quantized model
+gptq: bool | None
+# optional overrides to the bnb 4bit quantization configuration
+bnb_config_kwargs: dict[str, Any] | None
+
+# loraplus learning rate ratio lr_B / lr_A. Recommended value is 2^4.
+loraplus_lr_ratio: float | None
+# loraplus learning rate for lora embedding layers. Default value is 1e-6.
+loraplus_lr_embedding: float | None = 1e-06
+
+merge_lora: bool | None
+
+# Whether to use ReLoRA. Use with jagged_restart_*steps options.
+relora: bool | None
+# threshold for optimizer magnitude when pruning
+relora_prune_ratio: float | None
+# True to perform lora weight merges on cpu during restarts, for modest gpu memory
+# savings
+relora_cpu_offload: bool | None
+
+# how often to reset for jagged restarts
+jagged_restart_steps: int | None
+# how many warmup steps to take after reset for jagged restarts
+jagged_restart_warmup_steps: int | None
+# how many anneal steps to take before reset for jagged restarts
+jagged_restart_anneal_steps: int | None
+
+# If greater than 1, backpropagation will be skipped and the gradients will be
+# accumulated for the given number of steps.
+gradient_accumulation_steps: int | None = 1
+# The number of samples to include in each batch. This is the number of samples sent to
+# each GPU. Batch size per gpu = micro_batch_size * gradient_accumulation_steps
+micro_batch_size: int | None = 1
+# Total batch size, we do not recommended setting this manually
+batch_size: int | None
+# per gpu micro batch size for evals, defaults to value of micro_batch_size
+eval_batch_size: int | None
+
+# whether to find batch size that fits in memory. Passed to underlying transformers
+# Trainer
+auto_find_batch_size: bool | None
-# whether to find batch size that fits in memory. Passed to underlying transformers
-# Trainer
-auto_find_batch_size: bool | None
-
-# Whether to mask out or include the human's prompt from the training labels
-train_on_inputs: bool | None = False
-# Group similarly sized data to minimize padding. May be slower to start, as it must
-# download and sort the entire dataset. Note that training loss may have an oscillating
-# pattern with this enabled.
-group_by_length: bool | None
-
-learning_rate: str | float (required)
-embedding_lr: float | None
-embedding_lr_scale: float | None
-# Specify weight decay
-weight_decay: float | None = 0.0
-# Specify optimizer
-optimizer: OptimizerNames | CustomSupportedOptimizers | None = OptimizerNames.ADAMW_TORCH_FUSED
-# Dictionary of arguments to pass to the optimizer
-optim_args: str | dict[str, Any] | None
-# The target modules to optimize, i.e. the module names that you would like to train,
-# right now this is used only for GaLore algorithm
-optim_target_modules: list[str] | Literal['all_linear'] | None
-# Path to torch distx for optim 'adamw_anyprecision'
-torchdistx_path: str | None
-lr_scheduler: SchedulerType | Literal['one_cycle'] | Literal['rex'] | None = SchedulerType.COSINE
-# Specify a scheduler and kwargs to use with the optimizer
-lr_scheduler_kwargs: dict[str, Any] | None
-lr_quadratic_warmup: bool | None
-# decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of
-# peak lr
-cosine_min_lr_ratio: float | None
-# freeze lr at some percentage of the step, e.g. cosine_constant_lr_ratio=0.8 means
-# start cosine_min_lr at 80% of training step
-cosine_constant_lr_ratio: float | None
-# Learning rate div factor
-lr_div_factor: float | None
-
-lr_groups: list[LrGroup] | None
- # For LrGroup:
- name: str (required)
- modules: list[str] (required)
- lr: float (required)
-
+# Whether to mask out or include the human's prompt from the training labels
+train_on_inputs: bool | None = False
+# Group similarly sized data to minimize padding. May be slower to start, as it must
+# download and sort the entire dataset. Note that training loss may have an oscillating
+# pattern with this enabled.
+group_by_length: bool | None
+
+learning_rate: str | float (required)
+embedding_lr: float | None
+embedding_lr_scale: float | None
+# Specify weight decay
+weight_decay: float | None = 0.0
+# Specify optimizer
+optimizer: OptimizerNames | CustomSupportedOptimizers | None = OptimizerNames.ADAMW_TORCH_FUSED
+# Dictionary of arguments to pass to the optimizer
+optim_args: str | dict[str, Any] | None
+# The target modules to optimize, i.e. the module names that you would like to train,
+# right now this is used only for GaLore algorithm
+optim_target_modules: list[str] | Literal['all_linear'] | None
+# Path to torch distx for optim 'adamw_anyprecision'
+torchdistx_path: str | None
+lr_scheduler: SchedulerType | Literal['one_cycle'] | Literal['rex'] | None = SchedulerType.COSINE
+# Specify a scheduler and kwargs to use with the optimizer
+lr_scheduler_kwargs: dict[str, Any] | None
+lr_quadratic_warmup: bool | None
+# decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of
+# peak lr
+cosine_min_lr_ratio: float | None
+# freeze lr at some percentage of the step, e.g. cosine_constant_lr_ratio=0.8 means
+# start cosine_min_lr at 80% of training step
+cosine_constant_lr_ratio: float | None
+# Learning rate div factor
+lr_div_factor: float | None
+
+lr_groups: list[LrGroup] | None
+ # For LrGroup:
+ name: str (required)
+ modules: list[str] (required)
+ lr: float (required)
+
+# adamw hyperparams
+adam_epsilon: float | None
+# only used for CAME Optimizer
+adam_epsilon2: float | None
# adamw hyperparams
-adam_epsilon: float | None
-# only used for CAME Optimizer
-adam_epsilon2: float | None
-# adamw hyperparams
-adam_beta1: float | None
-# adamw hyperparams
-adam_beta2: float | None
-# only used for CAME Optimizer
-adam_beta3: float | None
-
-# Dion Optimizer learning rate
-dion_lr: float | None
-# Dion Optimizer momentum
-dion_momentum: float | None
-# Dion Optimizer: r/d fraction for low-rank approximation. Used to compute the low-rank
-# dimension.
-dion_rank_fraction: float | None = 1.0
-# Dion Optimizer: Round up the low-rank dimension to a multiple of this number. This may
-# be useful to ensure even sharding.
-dion_rank_multiple_of: int | None = 1
+adam_beta1: float | None
+# adamw hyperparams
+adam_beta2: float | None
+# only used for CAME Optimizer
+adam_beta3: float | None
+
+# Dion Optimizer learning rate
+dion_lr: float | None
+# Dion Optimizer momentum
+dion_momentum: float | None
+# Dion Optimizer: r/d fraction for low-rank approximation. Used to compute the low-rank
+# dimension.
+dion_rank_fraction: float | None = 1.0
+# Dion Optimizer: Round up the low-rank dimension to a multiple of this number. This may
+# be useful to ensure even sharding.
+dion_rank_multiple_of: int | None = 1
+
+# Gradient clipping max norm
+max_grad_norm: float | None
+num_epochs: float = 1.0
-# Gradient clipping max norm
-max_grad_norm: float | None
-num_epochs: float = 1.0
-
-use_wandb: bool | None
-# Set the name of your wandb run
-wandb_name: str | None
-# Set the ID of your wandb run
-wandb_run_id: str | None
-# "offline" to save run metadata locally and not sync to the server, "disabled" to turn
-# off wandb
-wandb_mode: str | None
-# Your wandb project name
-wandb_project: str | None
-# A wandb Team name if using a Team
-wandb_entity: str | None
-wandb_watch: str | None
-# "checkpoint" to log model to wandb Artifacts every `save_steps` or "end" to log only
-# at the end of training
-wandb_log_model: str | None
-
-use_mlflow: bool | None
-# URI to mlflow
-mlflow_tracking_uri: str | None
-# Your experiment name
-mlflow_experiment_name: str | None
-# Your run name
-mlflow_run_name: str | None
-# set to true to copy each saved checkpoint on each save to mlflow artifact registry
-hf_mlflow_log_artifacts: bool | None
-
-# Enable or disable Comet integration.
-use_comet: bool | None
-# API key for Comet. Recommended to set via `comet login`.
-comet_api_key: str | None
-# Workspace name in Comet. Defaults to the user's default workspace.
-comet_workspace: str | None
-# Project name in Comet. Defaults to Uncategorized.
-comet_project_name: str | None
-# Identifier for the experiment. Used to append data to an existing experiment or
-# control the key of new experiments. Default to a random key.
-comet_experiment_key: str | None
-# Create a new experiment ("create") or log to an existing one ("get"). Default
-# ("get_or_create") auto-selects based on configuration.
-comet_mode: str | None
-# Set to True to log data to Comet server, or False for offline storage. Default is
-# True.
-comet_online: bool | None
-# Dictionary for additional configuration settings, see the doc for more details.
-comet_experiment_config: dict[str, Any] | None
-
-# the number of activate layers in LISA
-lisa_n_layers: int | None
-# how often to switch layers in LISA
-lisa_step_interval: int | None
-# path under the model to access the layers
-lisa_layers_attribute: str | None = model.layers
-
-gradio_title: str | None
-gradio_share: bool | None
-gradio_server_name: str | None
-gradio_server_port: int | None
-gradio_max_new_tokens: int | None
-gradio_temperature: float | None
-
-use_ray: bool = False
-ray_run_name: str | None
-ray_num_workers: int = 1
-resources_per_worker: dict
-
-# The size of the image to resize to. It can be an integer (resized into padded-square
-# image) or a tuple (width, height).If not provided, we will attempt to load from
-# preprocessor.size, otherwise, images won't be resized.
-image_size: int | tuple[int, int] | None
-# The resampling algorithm to use for image resizing. Default is bilinear. Please refer
-# to PIL.Image.Resampling for more details.
-image_resize_algorithm: Literal['bilinear', 'bicubic', 'lanczos'] | Resampling | None
-
-# optional overrides to the base model configuration
-overrides_of_model_config: dict[str, Any] | None
-# optional overrides the base model loading from_pretrained
-overrides_of_model_kwargs: dict[str, Any] | None
-# If you want to specify the type of model to load, AutoModelForCausalLM is a good
-# choice too
-type_of_model: str | None
-# You can specify to choose a specific model revision from huggingface hub
-revision_of_model: str | None
-
-max_packed_sequence_len: int | None
-rope_scaling: Any | None
-noisy_embedding_alpha: float | None
-dpo_beta: float | None
-evaluation_strategy: str | None
+use_wandb: bool | None
+# Set the name of your wandb run
+wandb_name: str | None
+# Set the ID of your wandb run
+wandb_run_id: str | None
+# "offline" to save run metadata locally and not sync to the server, "disabled" to turn
+# off wandb
+wandb_mode: str | None
+# Your wandb project name
+wandb_project: str | None
+# A wandb Team name if using a Team
+wandb_entity: str | None
+wandb_watch: str | None
+# "checkpoint" to log model to wandb Artifacts every `save_steps` or "end" to log only
+# at the end of training
+wandb_log_model: str | None
+
+use_mlflow: bool | None
+# URI to mlflow
+mlflow_tracking_uri: str | None
+# Your experiment name
+mlflow_experiment_name: str | None
+# Your run name
+mlflow_run_name: str | None
+# set to true to copy each saved checkpoint on each save to mlflow artifact registry
+hf_mlflow_log_artifacts: bool | None
+
+# Enable or disable Comet integration.
+use_comet: bool | None
+# API key for Comet. Recommended to set via `comet login`.
+comet_api_key: str | None
+# Workspace name in Comet. Defaults to the user's default workspace.
+comet_workspace: str | None
+# Project name in Comet. Defaults to Uncategorized.
+comet_project_name: str | None
+# Identifier for the experiment. Used to append data to an existing experiment or
+# control the key of new experiments. Default to a random key.
+comet_experiment_key: str | None
+# Create a new experiment ("create") or log to an existing one ("get"). Default
+# ("get_or_create") auto-selects based on configuration.
+comet_mode: str | None
+# Set to True to log data to Comet server, or False for offline storage. Default is
+# True.
+comet_online: bool | None
+# Dictionary for additional configuration settings, see the doc for more details.
+comet_experiment_config: dict[str, Any] | None
+
+# the number of activate layers in LISA
+lisa_n_layers: int | None
+# how often to switch layers in LISA
+lisa_step_interval: int | None
+# path under the model to access the layers
+lisa_layers_attribute: str | None = model.layers
+
+gradio_title: str | None
+gradio_share: bool | None
+gradio_server_name: str | None
+gradio_server_port: int | None
+gradio_max_new_tokens: int | None
+gradio_temperature: float | None
+
+use_ray: bool = False
+ray_run_name: str | None
+ray_num_workers: int = 1
+resources_per_worker: dict
+
+# The size of the image to resize to. It can be an integer (resized into padded-square
+# image) or a tuple (width, height).If not provided, we will attempt to load from
+# preprocessor.size, otherwise, images won't be resized.
+image_size: int | tuple[int, int] | None
+# The resampling algorithm to use for image resizing. Default is bilinear. Please refer
+# to PIL.Image.Resampling for more details.
+image_resize_algorithm: Literal['bilinear', 'bicubic', 'lanczos'] | Resampling | None
+
+# optional overrides to the base model configuration
+overrides_of_model_config: dict[str, Any] | None
+# optional overrides the base model loading from_pretrained
+overrides_of_model_kwargs: dict[str, Any] | None
+# If you want to specify the type of model to load, AutoModelForCausalLM is a good
+# choice too
+type_of_model: str | None
+# You can specify to choose a specific model revision from huggingface hub
+revision_of_model: str | None
+
+max_packed_sequence_len: int | None
+rope_scaling: Any | None
+noisy_embedding_alpha: float | None
+dpo_beta: float | None
+evaluation_strategy: str | None
diff --git a/docs/quantize.html b/docs/quantize.html
index ecc611208..4b1ab6245 100644
--- a/docs/quantize.html
+++ b/docs/quantize.html
@@ -561,6 +561,20 @@ you used to train the model:
output_dir: # The path to the output directory used during training where the final checkpoint has been saved.
axolotl quantize qat.yml
This ensures that an identical quantization configuration is used to quantize the model as was used to train it.
+
+
+
+
+Note
+
+
+
+If you have configured pushing to hub with hub_model_id, your model hub name will have the quantization schema appended to it,
+e.g. axolotl-ai-cloud/qat-nvfp4-llama3B will become axolotl-ai-cloud/qat-nvfp4-llama3B-nvfp4w
+
+
diff --git a/index.html b/index.html
index 58cce67ea..6b1a0d2e8 100644
--- a/index.html
+++ b/index.html
@@ -499,7 +499,7 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
- 🎉 Latest Updates
- ✨ Overview
- - 🚀 Quick Start
+
- 🚀 Quick Start - LLM Fine-tuning in Minutes
- Google Colab
- Installation
@@ -528,6 +528,9 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
+A Free and Open Source LLM Fine-tuning Framework
+
+
@@ -573,19 +576,20 @@ Expand older updates
✨ Overview
-Axolotl is a tool designed to streamline post-training for various AI models.
+Axolotl is a free and open-source tool designed to streamline post-training and fine-tuning for the latest large language models (LLMs).
Features:
-- Multiple Model Support: Train various models like LLaMA, Mistral, Mixtral, Pythia, and more. We are compatible with HuggingFace transformers causal language models.
-- Training Methods: Full fine-tuning, LoRA, QLoRA, GPTQ, QAT, Preference Tuning (DPO, IPO, KTO, ORPO), RL (GRPO), Multimodal, and Reward Modelling (RM) / Process Reward Modelling (PRM).
-- Easy Configuration: Re-use a single YAML file between dataset preprocess, training, evaluation, quantization, and inference.
+- Multiple Model Support: Train various models like GPT-OSS, LLaMA, Mistral, Mixtral, Pythia, and many more models available on the Hugging Face Hub.
+- Multimodal Training: Fine-tune vision-language models (VLMs) including LLaMA-Vision, Qwen2-VL, Pixtral, LLaVA, SmolVLM2, and audio models like Voxtral with image, video, and audio support.
+- Training Methods: Full fine-tuning, LoRA, QLoRA, GPTQ, QAT, Preference Tuning (DPO, IPO, KTO, ORPO), RL (GRPO), and Reward Modelling (RM) / Process Reward Modelling (PRM).
+- Easy Configuration: Re-use a single YAML configuration file across the full fine-tuning pipeline: dataset preprocessing, training, evaluation, quantization, and inference.
- Performance Optimizations: Multipacking, Flash Attention, Xformers, Flex Attention, Liger Kernel, Cut Cross Entropy, Sequence Parallelism (SP), LoRA optimizations, Multi-GPU training (FSDP1, FSDP2, DeepSpeed), Multi-node training (Torchrun, Ray), and many more!
- Flexible Dataset Handling: Load from local, HuggingFace, and cloud (S3, Azure, GCP, OCI) datasets.
- Cloud Ready: We ship Docker images and also PyPI packages for use on cloud platforms and local hardware.
-
-🚀 Quick Start
+
+🚀 Quick Start - LLM Fine-tuning in Minutes
Requirements:
- NVIDIA GPU (Ampere or newer for
bf16 and Flash Attention) or AMD GPU
@@ -681,7 +685,7 @@ Expand older updates
📝 Citing Axolotl
If you use Axolotl in your research or projects, please cite it as follows:
@software{axolotl,
- title = {Axolotl: Post-Training for AI Models},
+ title = {Axolotl: Open Source LLM Post-Training},
author = {{Axolotl maintainers and contributors}},
url = {https://github.com/axolotl-ai-cloud/axolotl},
license = {Apache-2.0},
diff --git a/search.json b/search.json
index 8781fab7e..92bae473e 100644
--- a/search.json
+++ b/search.json
@@ -208,7 +208,7 @@
"href": "docs/quantize.html#configuring-quantization-in-axolotl",
"title": "Quantization with torchao",
"section": "Configuring Quantization in Axolotl",
- "text": "Configuring Quantization in Axolotl\nQuantization is configured using the quantization key in your configuration file.\nbase_model: # The path to the model to quantize.\nquantization:\n weight_dtype: # Optional[str] = \"int8\". Fake quantization layout to use for weight quantization. Valid options are uintX for X in [1, 2, 3, 4, 5, 6, 7], or int4, or int8\n activation_dtype: # Optional[str] = \"int8\". Fake quantization layout to use for activation quantization. Valid options are \"int4\" and \"int8\"\n group_size: # Optional[int] = 32. The number of elements in each group for per-group fake quantization\n quantize_embedding: # Optional[bool] = False. Whether to quantize the embedding layer.\n\noutput_dir: # The path to the output directory.\nOnce quantization is complete, your quantized model will be saved in the {output_dir}/quantized directory.\nYou may also use the quantize command to quantize a model which has been trained with QAT - you can do this by using the existing QAT configuration file which\nyou used to train the model:\n# qat.yml\nqat:\n activation_dtype: int8\n weight_dtype: int8\n group_size: 256\n quantize_embedding: true\n\noutput_dir: # The path to the output directory used during training where the final checkpoint has been saved.\naxolotl quantize qat.yml\nThis ensures that an identical quantization configuration is used to quantize the model as was used to train it.",
+ "text": "Configuring Quantization in Axolotl\nQuantization is configured using the quantization key in your configuration file.\nbase_model: # The path to the model to quantize.\nquantization:\n weight_dtype: # Optional[str] = \"int8\". Fake quantization layout to use for weight quantization. Valid options are uintX for X in [1, 2, 3, 4, 5, 6, 7], or int4, or int8\n activation_dtype: # Optional[str] = \"int8\". Fake quantization layout to use for activation quantization. Valid options are \"int4\" and \"int8\"\n group_size: # Optional[int] = 32. The number of elements in each group for per-group fake quantization\n quantize_embedding: # Optional[bool] = False. Whether to quantize the embedding layer.\n\noutput_dir: # The path to the output directory.\nOnce quantization is complete, your quantized model will be saved in the {output_dir}/quantized directory.\nYou may also use the quantize command to quantize a model which has been trained with QAT - you can do this by using the existing QAT configuration file which\nyou used to train the model:\n# qat.yml\nqat:\n activation_dtype: int8\n weight_dtype: int8\n group_size: 256\n quantize_embedding: true\n\noutput_dir: # The path to the output directory used during training where the final checkpoint has been saved.\naxolotl quantize qat.yml\nThis ensures that an identical quantization configuration is used to quantize the model as was used to train it.\n\n\n\n\n\n\nNote\n\n\n\nIf you have configured pushing to hub with hub_model_id, your model hub name will have the quantization schema appended to it,\ne.g. axolotl-ai-cloud/qat-nvfp4-llama3B will become axolotl-ai-cloud/qat-nvfp4-llama3B-nvfp4w",
"crumbs": [
"How To Guides",
"Quantization with torchao"
@@ -774,14 +774,14 @@
"href": "docs/api/cli.args.html",
"title": "cli.args",
"section": "",
- "text": "cli.args\nModule for axolotl CLI command arguments.\n\n\n\n\n\nName\nDescription\n\n\n\n\nEvaluateCliArgs\nDataclass with CLI arguments for axolotl evaluate command.\n\n\nInferenceCliArgs\nDataclass with CLI arguments for axolotl inference command.\n\n\nPreprocessCliArgs\nDataclass with CLI arguments for axolotl preprocess command.\n\n\nQuantizeCliArgs\nDataclass with CLI arguments for axolotl quantize command.\n\n\nTrainerCliArgs\nDataclass with CLI arguments for axolotl train command.\n\n\nVllmServeCliArgs\nDataclass with CLI arguments for axolotl vllm-serve command.\n\n\n\n\n\ncli.args.EvaluateCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n)\nDataclass with CLI arguments for axolotl evaluate command.\n\n\n\ncli.args.InferenceCliArgs(prompter=None)\nDataclass with CLI arguments for axolotl inference command.\n\n\n\ncli.args.PreprocessCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=1,\n prompter=None,\n download=True,\n iterable=False,\n)\nDataclass with CLI arguments for axolotl preprocess command.\n\n\n\ncli.args.QuantizeCliArgs(\n base_model=None,\n weight_dtype=None,\n activation_dtype=None,\n quantize_embedding=None,\n group_size=None,\n output_dir=None,\n)\nDataclass with CLI arguments for axolotl quantize command.\n\n\n\ncli.args.TrainerCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n prompter=None,\n shard=False,\n)\nDataclass with CLI arguments for axolotl train command.\n\n\n\ncli.args.VllmServeCliArgs(\n tensor_parallel_size=None,\n data_parallel_size=None,\n host=None,\n port=None,\n gpu_memory_utilization=None,\n dtype=None,\n max_model_len=None,\n enable_prefix_caching=None,\n serve_module=None,\n enable_reasoning=None,\n reasoning_parser=None,\n)\nDataclass with CLI arguments for axolotl vllm-serve command."
+ "text": "cli.args\nModule for axolotl CLI command arguments.\n\n\n\n\n\nName\nDescription\n\n\n\n\nEvaluateCliArgs\nDataclass with CLI arguments for axolotl evaluate command.\n\n\nInferenceCliArgs\nDataclass with CLI arguments for axolotl inference command.\n\n\nPreprocessCliArgs\nDataclass with CLI arguments for axolotl preprocess command.\n\n\nQuantizeCliArgs\nDataclass with CLI arguments for axolotl quantize command.\n\n\nTrainerCliArgs\nDataclass with CLI arguments for axolotl train command.\n\n\nVllmServeCliArgs\nDataclass with CLI arguments for axolotl vllm-serve command.\n\n\n\n\n\ncli.args.EvaluateCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n)\nDataclass with CLI arguments for axolotl evaluate command.\n\n\n\ncli.args.InferenceCliArgs(prompter=None)\nDataclass with CLI arguments for axolotl inference command.\n\n\n\ncli.args.PreprocessCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=1,\n prompter=None,\n download=True,\n iterable=False,\n)\nDataclass with CLI arguments for axolotl preprocess command.\n\n\n\ncli.args.QuantizeCliArgs(\n base_model=None,\n weight_dtype=None,\n activation_dtype=None,\n quantize_embedding=None,\n group_size=None,\n output_dir=None,\n hub_model_id=None,\n)\nDataclass with CLI arguments for axolotl quantize command.\n\n\n\ncli.args.TrainerCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n prompter=None,\n shard=False,\n)\nDataclass with CLI arguments for axolotl train command.\n\n\n\ncli.args.VllmServeCliArgs(\n tensor_parallel_size=None,\n data_parallel_size=None,\n host=None,\n port=None,\n gpu_memory_utilization=None,\n dtype=None,\n max_model_len=None,\n enable_prefix_caching=None,\n serve_module=None,\n enable_reasoning=None,\n reasoning_parser=None,\n)\nDataclass with CLI arguments for axolotl vllm-serve command."
},
{
"objectID": "docs/api/cli.args.html#classes",
"href": "docs/api/cli.args.html#classes",
"title": "cli.args",
"section": "",
- "text": "Name\nDescription\n\n\n\n\nEvaluateCliArgs\nDataclass with CLI arguments for axolotl evaluate command.\n\n\nInferenceCliArgs\nDataclass with CLI arguments for axolotl inference command.\n\n\nPreprocessCliArgs\nDataclass with CLI arguments for axolotl preprocess command.\n\n\nQuantizeCliArgs\nDataclass with CLI arguments for axolotl quantize command.\n\n\nTrainerCliArgs\nDataclass with CLI arguments for axolotl train command.\n\n\nVllmServeCliArgs\nDataclass with CLI arguments for axolotl vllm-serve command.\n\n\n\n\n\ncli.args.EvaluateCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n)\nDataclass with CLI arguments for axolotl evaluate command.\n\n\n\ncli.args.InferenceCliArgs(prompter=None)\nDataclass with CLI arguments for axolotl inference command.\n\n\n\ncli.args.PreprocessCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=1,\n prompter=None,\n download=True,\n iterable=False,\n)\nDataclass with CLI arguments for axolotl preprocess command.\n\n\n\ncli.args.QuantizeCliArgs(\n base_model=None,\n weight_dtype=None,\n activation_dtype=None,\n quantize_embedding=None,\n group_size=None,\n output_dir=None,\n)\nDataclass with CLI arguments for axolotl quantize command.\n\n\n\ncli.args.TrainerCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n prompter=None,\n shard=False,\n)\nDataclass with CLI arguments for axolotl train command.\n\n\n\ncli.args.VllmServeCliArgs(\n tensor_parallel_size=None,\n data_parallel_size=None,\n host=None,\n port=None,\n gpu_memory_utilization=None,\n dtype=None,\n max_model_len=None,\n enable_prefix_caching=None,\n serve_module=None,\n enable_reasoning=None,\n reasoning_parser=None,\n)\nDataclass with CLI arguments for axolotl vllm-serve command."
+ "text": "Name\nDescription\n\n\n\n\nEvaluateCliArgs\nDataclass with CLI arguments for axolotl evaluate command.\n\n\nInferenceCliArgs\nDataclass with CLI arguments for axolotl inference command.\n\n\nPreprocessCliArgs\nDataclass with CLI arguments for axolotl preprocess command.\n\n\nQuantizeCliArgs\nDataclass with CLI arguments for axolotl quantize command.\n\n\nTrainerCliArgs\nDataclass with CLI arguments for axolotl train command.\n\n\nVllmServeCliArgs\nDataclass with CLI arguments for axolotl vllm-serve command.\n\n\n\n\n\ncli.args.EvaluateCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n)\nDataclass with CLI arguments for axolotl evaluate command.\n\n\n\ncli.args.InferenceCliArgs(prompter=None)\nDataclass with CLI arguments for axolotl inference command.\n\n\n\ncli.args.PreprocessCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=1,\n prompter=None,\n download=True,\n iterable=False,\n)\nDataclass with CLI arguments for axolotl preprocess command.\n\n\n\ncli.args.QuantizeCliArgs(\n base_model=None,\n weight_dtype=None,\n activation_dtype=None,\n quantize_embedding=None,\n group_size=None,\n output_dir=None,\n hub_model_id=None,\n)\nDataclass with CLI arguments for axolotl quantize command.\n\n\n\ncli.args.TrainerCliArgs(\n debug=False,\n debug_text_only=False,\n debug_num_examples=0,\n prompter=None,\n shard=False,\n)\nDataclass with CLI arguments for axolotl train command.\n\n\n\ncli.args.VllmServeCliArgs(\n tensor_parallel_size=None,\n data_parallel_size=None,\n host=None,\n port=None,\n gpu_memory_utilization=None,\n dtype=None,\n max_model_len=None,\n enable_prefix_caching=None,\n serve_module=None,\n enable_reasoning=None,\n reasoning_parser=None,\n)\nDataclass with CLI arguments for axolotl vllm-serve command."
},
{
"objectID": "docs/api/monkeypatch.gradient_checkpointing.offload_cpu.html",
@@ -1110,14 +1110,14 @@
"href": "docs/api/utils.quantization.html",
"title": "utils.quantization",
"section": "",
- "text": "utils.quantization\nUtilities for quantization including QAT and PTQ using torchao.\n\n\n\n\n\nName\nDescription\n\n\n\n\nconvert_qat_model_for_ptq\nThis function is used to convert a swap fake-quantized modules in a model\n\n\nget_ptq_config\nThis function is used to build a post-training quantization config.\n\n\nprepare_model_for_qat\nThis function is used to prepare a model for QAT by swapping the model’s linear\n\n\nquantize_model_for_ptq\nThis function is used to quantize a model for post-training quantization.\n\n\n\n\n\nutils.quantization.convert_qat_model_for_ptq(model, *, quantize_embedding=None)\nThis function is used to convert a swap fake-quantized modules in a model\nwhich has been trained with QAT back to the original modules, ready for PTQ.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to convert.\nrequired\n\n\nquantize_embedding\nbool | None\nWhether to quantize the model’s embedding weights.\nNone\n\n\n\n\n\n\n\nutils.quantization.get_ptq_config(\n weight_dtype,\n activation_dtype=None,\n group_size=None,\n)\nThis function is used to build a post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nweight_dtype\nTorchIntDType\nThe dtype to use for weight quantization.\nrequired\n\n\nactivation_dtype\nTorchIntDType | None\nThe dtype to use for activation quantization.\nNone\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nAOBaseConfig\nThe post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation dtype is not specified and the weight dtype is not int8 or int4, or if the group size is not specified for int8 or int4 weight only quantization.\n\n\n\n\n\n\n\nutils.quantization.prepare_model_for_qat(\n model,\n weight_dtype,\n group_size,\n activation_dtype=None,\n quantize_embedding=False,\n)\nThis function is used to prepare a model for QAT by swapping the model’s linear\nlayers with fake quantized linear layers, and optionally the embedding weights with\nfake quantized embedding weights.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchIntDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint\nThe group size to use for weight quantization.\nrequired\n\n\nactivation_dtype\nTorchIntDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool\nWhether to quantize the model’s embedding weights.\nFalse\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation/weight dtype combination is invalid.\n\n\n\n\n\n\n\nutils.quantization.quantize_model_for_ptq(\n model,\n weight_dtype,\n group_size=None,\n activation_dtype=None,\n quantize_embedding=None,\n)\nThis function is used to quantize a model for post-training quantization.\nIt swaps the model’s linear layers with fake quantized linear layers.\nIf quantize_embedding is True, it will also swap the model’s embedding weights with fake quantized embedding weights.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchIntDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\nactivation_dtype\nTorchIntDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool | None\nWhether to quantize the model’s embedding weights.\nNone"
+ "text": "utils.quantization\nUtilities for quantization including QAT and PTQ using torchao.\n\n\n\n\n\nName\nDescription\n\n\n\n\nconvert_qat_model\nThis function converts a QAT model which has fake quantized layers back to the original model.\n\n\nget_quantization_config\nThis function is used to build a post-training quantization config.\n\n\nprepare_model_for_qat\nThis function is used to prepare a model for QAT by swapping the model’s linear\n\n\nquantize_model\nThis function is used to quantize a model.\n\n\n\n\n\nutils.quantization.convert_qat_model(model, quantize_embedding=False)\nThis function converts a QAT model which has fake quantized layers back to the original model.\n\n\n\nutils.quantization.get_quantization_config(\n weight_dtype,\n activation_dtype=None,\n group_size=None,\n)\nThis function is used to build a post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nweight_dtype\nTorchAOQuantDType\nThe dtype to use for weight quantization.\nrequired\n\n\nactivation_dtype\nTorchAOQuantDType | None\nThe dtype to use for activation quantization.\nNone\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nAOBaseConfig\nThe post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation dtype is not specified and the weight dtype is not int8 or int4, or if the group size is not specified for int8 or int4 weight only quantization.\n\n\n\n\n\n\n\nutils.quantization.prepare_model_for_qat(\n model,\n weight_dtype,\n group_size=None,\n activation_dtype=None,\n quantize_embedding=False,\n)\nThis function is used to prepare a model for QAT by swapping the model’s linear\nlayers with fake quantized linear layers, and optionally the embedding weights with\nfake quantized embedding weights.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchAOQuantDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\nactivation_dtype\nTorchAOQuantDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool\nWhether to quantize the model’s embedding weights.\nFalse\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation/weight dtype combination is invalid.\n\n\n\n\n\n\n\nutils.quantization.quantize_model(\n model,\n weight_dtype,\n group_size=None,\n activation_dtype=None,\n quantize_embedding=None,\n)\nThis function is used to quantize a model.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchAOQuantDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\nactivation_dtype\nTorchAOQuantDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool | None\nWhether to quantize the model’s embedding weights.\nNone"
},
{
"objectID": "docs/api/utils.quantization.html#functions",
"href": "docs/api/utils.quantization.html#functions",
"title": "utils.quantization",
"section": "",
- "text": "Name\nDescription\n\n\n\n\nconvert_qat_model_for_ptq\nThis function is used to convert a swap fake-quantized modules in a model\n\n\nget_ptq_config\nThis function is used to build a post-training quantization config.\n\n\nprepare_model_for_qat\nThis function is used to prepare a model for QAT by swapping the model’s linear\n\n\nquantize_model_for_ptq\nThis function is used to quantize a model for post-training quantization.\n\n\n\n\n\nutils.quantization.convert_qat_model_for_ptq(model, *, quantize_embedding=None)\nThis function is used to convert a swap fake-quantized modules in a model\nwhich has been trained with QAT back to the original modules, ready for PTQ.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to convert.\nrequired\n\n\nquantize_embedding\nbool | None\nWhether to quantize the model’s embedding weights.\nNone\n\n\n\n\n\n\n\nutils.quantization.get_ptq_config(\n weight_dtype,\n activation_dtype=None,\n group_size=None,\n)\nThis function is used to build a post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nweight_dtype\nTorchIntDType\nThe dtype to use for weight quantization.\nrequired\n\n\nactivation_dtype\nTorchIntDType | None\nThe dtype to use for activation quantization.\nNone\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nAOBaseConfig\nThe post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation dtype is not specified and the weight dtype is not int8 or int4, or if the group size is not specified for int8 or int4 weight only quantization.\n\n\n\n\n\n\n\nutils.quantization.prepare_model_for_qat(\n model,\n weight_dtype,\n group_size,\n activation_dtype=None,\n quantize_embedding=False,\n)\nThis function is used to prepare a model for QAT by swapping the model’s linear\nlayers with fake quantized linear layers, and optionally the embedding weights with\nfake quantized embedding weights.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchIntDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint\nThe group size to use for weight quantization.\nrequired\n\n\nactivation_dtype\nTorchIntDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool\nWhether to quantize the model’s embedding weights.\nFalse\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation/weight dtype combination is invalid.\n\n\n\n\n\n\n\nutils.quantization.quantize_model_for_ptq(\n model,\n weight_dtype,\n group_size=None,\n activation_dtype=None,\n quantize_embedding=None,\n)\nThis function is used to quantize a model for post-training quantization.\nIt swaps the model’s linear layers with fake quantized linear layers.\nIf quantize_embedding is True, it will also swap the model’s embedding weights with fake quantized embedding weights.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchIntDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\nactivation_dtype\nTorchIntDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool | None\nWhether to quantize the model’s embedding weights.\nNone"
+ "text": "Name\nDescription\n\n\n\n\nconvert_qat_model\nThis function converts a QAT model which has fake quantized layers back to the original model.\n\n\nget_quantization_config\nThis function is used to build a post-training quantization config.\n\n\nprepare_model_for_qat\nThis function is used to prepare a model for QAT by swapping the model’s linear\n\n\nquantize_model\nThis function is used to quantize a model.\n\n\n\n\n\nutils.quantization.convert_qat_model(model, quantize_embedding=False)\nThis function converts a QAT model which has fake quantized layers back to the original model.\n\n\n\nutils.quantization.get_quantization_config(\n weight_dtype,\n activation_dtype=None,\n group_size=None,\n)\nThis function is used to build a post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nweight_dtype\nTorchAOQuantDType\nThe dtype to use for weight quantization.\nrequired\n\n\nactivation_dtype\nTorchAOQuantDType | None\nThe dtype to use for activation quantization.\nNone\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nAOBaseConfig\nThe post-training quantization config.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation dtype is not specified and the weight dtype is not int8 or int4, or if the group size is not specified for int8 or int4 weight only quantization.\n\n\n\n\n\n\n\nutils.quantization.prepare_model_for_qat(\n model,\n weight_dtype,\n group_size=None,\n activation_dtype=None,\n quantize_embedding=False,\n)\nThis function is used to prepare a model for QAT by swapping the model’s linear\nlayers with fake quantized linear layers, and optionally the embedding weights with\nfake quantized embedding weights.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchAOQuantDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\nactivation_dtype\nTorchAOQuantDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool\nWhether to quantize the model’s embedding weights.\nFalse\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\n\n\n\n\n\nValueError\nIf the activation/weight dtype combination is invalid.\n\n\n\n\n\n\n\nutils.quantization.quantize_model(\n model,\n weight_dtype,\n group_size=None,\n activation_dtype=None,\n quantize_embedding=None,\n)\nThis function is used to quantize a model.\n\n\n\n\n\n\n\n\n\n\n\nName\nType\nDescription\nDefault\n\n\n\n\nmodel\n\nThe model to quantize.\nrequired\n\n\nweight_dtype\nTorchAOQuantDType\nThe dtype to use for weight quantization.\nrequired\n\n\ngroup_size\nint | None\nThe group size to use for weight quantization.\nNone\n\n\nactivation_dtype\nTorchAOQuantDType | None\nThe dtype to use for activation quantization.\nNone\n\n\nquantize_embedding\nbool | None\nWhether to quantize the model’s embedding weights.\nNone"
},
{
"objectID": "docs/api/utils.dict.html",
@@ -1327,14 +1327,14 @@
"href": "docs/api/utils.schemas.enums.html",
"title": "utils.schemas.enums",
"section": "",
- "text": "utils.schemas.enums\nEnums for Axolotl input config\n\n\n\n\n\nName\nDescription\n\n\n\n\nChatTemplate\nChat templates configuration subset\n\n\nCustomSupportedOptimizers\nCustom supported optimizers\n\n\nRLType\nRL trainer type configuration subset\n\n\nRingAttnFunc\nEnum class for supported ring-flash-attn implementations\n\n\nTorchIntDType\nTorch integer data types - getattr guards against torch < 2.6 which does not support int4\n\n\n\n\n\nutils.schemas.enums.ChatTemplate()\nChat templates configuration subset\n\n\n\nutils.schemas.enums.CustomSupportedOptimizers()\nCustom supported optimizers\n\n\n\nutils.schemas.enums.RLType()\nRL trainer type configuration subset\n\n\n\nutils.schemas.enums.RingAttnFunc()\nEnum class for supported ring-flash-attn implementations\n\n\n\nutils.schemas.enums.TorchIntDType()\nTorch integer data types - getattr guards against torch < 2.6 which does not support int4"
+ "text": "utils.schemas.enums\nEnums for Axolotl input config\n\n\n\n\n\nName\nDescription\n\n\n\n\nChatTemplate\nChat templates configuration subset\n\n\nCustomSupportedOptimizers\nCustom supported optimizers\n\n\nRLType\nRL trainer type configuration subset\n\n\nRingAttnFunc\nEnum class for supported ring-flash-attn implementations\n\n\n\n\n\nutils.schemas.enums.ChatTemplate()\nChat templates configuration subset\n\n\n\nutils.schemas.enums.CustomSupportedOptimizers()\nCustom supported optimizers\n\n\n\nutils.schemas.enums.RLType()\nRL trainer type configuration subset\n\n\n\nutils.schemas.enums.RingAttnFunc()\nEnum class for supported ring-flash-attn implementations"
},
{
"objectID": "docs/api/utils.schemas.enums.html#classes",
"href": "docs/api/utils.schemas.enums.html#classes",
"title": "utils.schemas.enums",
"section": "",
- "text": "Name\nDescription\n\n\n\n\nChatTemplate\nChat templates configuration subset\n\n\nCustomSupportedOptimizers\nCustom supported optimizers\n\n\nRLType\nRL trainer type configuration subset\n\n\nRingAttnFunc\nEnum class for supported ring-flash-attn implementations\n\n\nTorchIntDType\nTorch integer data types - getattr guards against torch < 2.6 which does not support int4\n\n\n\n\n\nutils.schemas.enums.ChatTemplate()\nChat templates configuration subset\n\n\n\nutils.schemas.enums.CustomSupportedOptimizers()\nCustom supported optimizers\n\n\n\nutils.schemas.enums.RLType()\nRL trainer type configuration subset\n\n\n\nutils.schemas.enums.RingAttnFunc()\nEnum class for supported ring-flash-attn implementations\n\n\n\nutils.schemas.enums.TorchIntDType()\nTorch integer data types - getattr guards against torch < 2.6 which does not support int4"
+ "text": "Name\nDescription\n\n\n\n\nChatTemplate\nChat templates configuration subset\n\n\nCustomSupportedOptimizers\nCustom supported optimizers\n\n\nRLType\nRL trainer type configuration subset\n\n\nRingAttnFunc\nEnum class for supported ring-flash-attn implementations\n\n\n\n\n\nutils.schemas.enums.ChatTemplate()\nChat templates configuration subset\n\n\n\nutils.schemas.enums.CustomSupportedOptimizers()\nCustom supported optimizers\n\n\n\nutils.schemas.enums.RLType()\nRL trainer type configuration subset\n\n\n\nutils.schemas.enums.RingAttnFunc()\nEnum class for supported ring-flash-attn implementations"
},
{
"objectID": "docs/api/utils.callbacks.profiler.html",
@@ -3851,7 +3851,7 @@
"href": "docs/config-reference.html",
"title": "Config Reference",
"section": "",
- "text": "# Allow overwrite yml config using from cli\nstrict: bool | None = False\n# Resume from a specific checkpoint dir\nresume_from_checkpoint: str | None\n# If resume_from_checkpoint isn't set and you simply want it to start where it left off.\n# Be careful with this being turned on between different models.\nauto_resume_from_checkpoints: bool | None\n# Resize the model embeddings when new tokens are added to multiples of 32. This is\n# reported to improve training speed on some models\nresize_token_embeddings_to_32x: bool | None\nmean_resizing_embeddings: bool | None = False\n\n# Whether to shrink the embeddings to len(tokenizer). By default, we won't shrink.\nshrink_embeddings: bool | None\n# Don't upcast the embeddings to float32 when using PEFT. Useful for low-VRAM GPUs\nembeddings_skip_upcast: bool | None\n# Reinitialize model weights randomly instead of loading pretrained weights\nreinit_weights: bool | None\n\n# module to custom trainer class to use for training\ntrainer_cls: str | None\n\n# Use RL training: 'dpo', 'ipo', 'kto', 'simpo', 'orpo', 'grpo'\nrl: RLType | None\n\ntrl: TRLConfig | None\n # For TRLConfig:\n # Beta parameter for the RL training. Same as `rl_beta`. Use\n beta: float | None\n # Maximum length of the completion for RL training.\n max_completion_length: int | None\n\n # Whether to use VLLM for RL training.\n use_vllm: bool = False\n # VLLM mode to use, one of 'server' or 'colocate'\n vllm_mode: Literal['server', 'colocate'] | None\n # Host of the vLLM server to connect to.\n vllm_server_host: str | None = 0.0.0.0\n # Port of the vLLM server to connect to.\n vllm_server_port: int | None = 8000\n # Total timeout (in seconds) to wait for the vLLM server to respond.\n vllm_server_timeout: int | None\n # Regex for vLLM guided decoding.\n vllm_guided_decoding_regex: str | None\n\n # List of reward functions to load. Paths must be importable from current dir.\n reward_funcs: list[str] | None\n # List of reward weights for the reward functions.\n reward_weights: list[float] | None\n # Number of generations to sample.\n num_generations: int | None\n # Whether to log completions.\n log_completions: bool | None = False\n # Number of completions to print when log_completions is True.\n num_completions_to_print: int | None\n # Controls whether importance sampling ratios are computed at the `'token'` or\n # `'sequence'` level. For GSPO, use `sequence`, default is None which corresponds to\n # the original GRPO paper.\n importance_sampling_level: Literal['sequence', 'token'] | None\n\n # Whether to sync the reference model.\n sync_ref_model: bool | None = False\n # Mixup alpha for the reference model.\n ref_model_mixup_alpha: float | None = 0.9\n # Sync steps for the reference model.\n ref_model_sync_steps: int | None = 64\n # Whether to scale rewards by their standard deviation.\n scale_rewards: bool = True\n\n # Sampling temperature for the GRPO policy.\n temperature: float | None\n # Top-p sampling probability for the generation policy.\n top_p: float | None\n # Top-k sampling for the generation policy.\n top_k: int | None\n # Minimum probability for the generation policy.\n min_p: float | None\n # Penalty for tokens that appear in prompt and generated text.\n repetition_penalty: float | None\n # Number of iterations per batch (μ) for GRPO.\n num_iterations: int | None\n # Epsilon value for clipping in the GRPO algorithm.\n epsilon: float | None\n # Upper-bound epsilon value for clipping in the GRPO algorithm.\n epsilon_high: float | None\n # Whether to use Liger loss for GRPO.\n use_liger_loss: bool | None\n # Loss formulation to use. Supported values: grpo, bnpo, dr_grpo.\n loss_type: str | None\n # Whether to exclude truncated completions from loss calculation.\n mask_truncated_completions: bool = False\n\nvllm: VllmConfig | None\n # For VllmConfig:\n # Device to use for VLLM\n device: str | None = auto\n # Tensor parallel size for VLLM\n tensor_parallel_size: int | None\n # Data parallel size for VLLM\n data_parallel_size: int | None\n # GPU memory utilization for VLLM\n gpu_memory_utilization: float | None = 0.9\n # Data type for VLLM\n dtype: str | None = auto\n # Maximum length of the model context for VLLM\n max_model_len: int | None\n # Enable prefix caching for VLLM\n enable_prefix_caching: bool | None\n # Host for the vLLM server to start on\n host: str | None = 0.0.0.0\n # Port of the vLLM server to start on\n port: int | None = 8000\n\n # Enable reasoning for VLLM\n enable_reasoning: bool | None\n # Reasoning parser for VLLM\n reasoning_parser: str | None\n\nqat: QATConfig | None\n # For QATConfig:\n # Fake quantization layout to use for activation quantization. Valid options are\n # \"int4\" and \"int8\"\n activation_dtype: TorchIntDType | None\n # Fake quantization layout to use for weight quantization. Valid options are \"int4\"\n # and \"int8\"\n weight_dtype: TorchIntDType = TorchIntDType.int8\n # Quantize embedding\n quantize_embedding: bool | None = False\n # The number of elements in each group for per-group fake quantization\n group_size: int | None = 32\n # The number of steps to apply fake quantization after\n fake_quant_after_n_steps: int | None\n\nquantization: PTQConfig | None\n # For PTQConfig:\n # Fake quantization layout to use for weight quantization. Valid options are uintX for\n # X in [1, 2, 3, 4, 5, 6, 7], or int4, or int8\n weight_dtype: TorchIntDType = TorchIntDType.int8\n # Fake quantization layout to use for activation quantization. Valid options are\n # \"int4\" and \"int8\"\n activation_dtype: TorchIntDType | None\n # Whether to quantize the embedding layer.\n quantize_embedding: bool | None\n # The number of elements in each group for per-group fake quantization\n group_size: int | None = 32\n\n# Reward modelling: `True` or `False`\nreward_model: bool | None\n# Process reward modelling: `True` or `False`\nprocess_reward_model: bool | None\n# Coefficient to incentivize the reward model to output mean-zero rewards (proposed by\n# https://huggingface.co/papers/2312.09244, Eq. 2). Recommended value: `0.01`.\ncenter_rewards_coefficient: float | None\nnum_labels: int | None\n\n# Whether to perform weighting in DPO trainer\ndpo_use_weighting: bool | None\ndpo_use_logits_to_keep: bool | None\ndpo_label_smoothing: float | None\ndpo_norm_loss: bool | None\ndpo_padding_free: bool | None\ndpo_generate_during_eval: bool | None\n\n# A list of one or more datasets to finetune the model with\ndatasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None\n # For SFTDataset:\n # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory\n path: str | None\n # name of dataset split to load from\n split: str | None\n # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]\n type: str | UserDefinedPrompterType | None\n # For UserDefinedPrompterType:\n # Custom user instruction prompt\n system_prompt: str | None\n # Use {system} as key to be replaced\n system_format: str | None\n field_system: str | None\n field_instruction: str | None\n field_input: str | None\n field_output: str | None\n\n # Customizable to be single line or multi-line. Use {instruction}/{input} as key to\n # be replaced. 'format' can include {input}\n format: str | None\n # 'no_input_format' cannot include {input}\n no_input_format: str | None\n input_transform: str | None\n # split dataset into N pieces (use with shards_idx)\n shards: int | None\n # the index of sharded dataset to use\n shards_idx: int | None\n # process dataset in N sequential chunks for memory efficiency (exclusive with\n # `shards`)\n preprocess_shards: int | None\n conversation: str | None\n\n # The name of the chat template to use for training, following values are supported:\n # tokenizer_default: Uses the chat template that is available in the\n # tokenizer_config.json. If the chat template is not available in the tokenizer, it\n # will raise an error. This is the default.\n # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n # tokenizer_default_fallback_*: where * is the name of the chat template to fallback\n # to if the tokenizer does not have a chat template else default to tokenizer. E.g.\n # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat\n # template. The custom jinja template should be provided in the chat_template_jinja\n # field.\n chat_template: ChatTemplate | str | None\n # Custom jinja chat template or path to jinja file. Used only if `chat_template:\n # jinja` or empty.\n chat_template_jinja: str | None\n # path to source data files\n data_files: str | list[str] | None\n input_format: str | None\n # name of dataset configuration to load\n name: str | None\n # defines the datatype when path is a file\n ds_type: str | None\n # For `completion` datasets only, uses the provided field instead of `text` column\n field: str | None\n field_human: str | None\n field_model: str | None\n # Key containing the messages (default: \"messages\")\n field_messages: str | None\n # Key containing the tools (default: \"tools\"). Must be a list[dict] and follow [JSON\n # schema](https://json-schema.org/learn/getting-started-step-by-step).\n field_tools: str | None\n # Key containing the reasoning trace (default: \"reasoning_content\").\n field_thinking: str | None\n # The key the chat template expects that indicates the reasoning trace.\n template_thinking_key: str | None\n\n message_field_role: str | None\n\n message_field_content: str | None\n # Mapping of properties from the input dataset to the chat template. (default:\n # message_property_mappings={'role':'role', 'content':'content'}) If a property exists\n # in the template but not in this mapping, the system will attempt to load it directly\n # from the message using the property name as the key. Example: In the mapping below,\n # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and\n # used as 'content' in the chat template.\n message_property_mappings: dict[str, str] | None\n # The key in the message turn that indicates via boolean whether tokens of a turn\n # should be considered for training. Useful to selectively train on certain turns\n # besides the `roles_to_train`.\n message_field_training: str | None\n # The key in the message turn that contains the training details. Useful to\n # selectively train on certain tokens in a turn. The value of the key is a List[Dict]\n # containing `begin_offset` (start character index in content), `end_offset` (end\n # character index in content), and `train` (boolean whether to train).\n message_field_training_detail: str | None\n # (for Qwen3 template only) Whether to split the assistant content based on a\n # reasoning trace inside delimited tags\n split_thinking: bool | None\n logprobs_field: str | None\n temperature: float | None\n # Roles to train on. The tokens from these roles will be considered for the loss.\n roles_to_train: list[str] | None\n # Which EOS tokens to train on in the conversation. Possible values are: all: train on\n # all EOS tokens, turn (default): train on the EOS token at the end of each trainable\n # turn, last: train on the last EOS token in the conversation\n train_on_eos: Literal['all', 'turn', 'last'] | None\n # Roles mapping in the messages. The format is {target_role: [source_roles]}. All\n # source roles will be mapped to the target role. The default is: user: [\"human\",\n # \"user\"], assistant: [\"gpt\", \"assistant\"], system: [\"system\"], tool: [\"tool\"]\n roles: dict[str, list[str]] | None\n # Whether to drop the system turn from the dataset. Only works with chat_template.\n # This does not drop the default system message from chat_template if it exists. If\n # you wish to, we recommend using a custom jinja template with the default system\n # message removed or adding a system turn with empty content.\n drop_system_message: bool | None\n # Trust remote code for untrusted source\n trust_remote_code: bool | None = False\n # The specific revision of the dataset to use when loading from the Hugging Face Hub.\n # This can be a commit hash, tag, or branch name. If not specified, the latest version\n # will be used. This parameter is ignored for local datasets.\n revision: str | None\n\n # For DPODataset:\n path: str | None\n split: str | None\n type: UserDefinedDPOType | str | None\n # For UserDefinedDPOType:\n field_system: str | None\n field_prompt: str | None\n field_chosen: str | None\n field_rejected: str | None\n prompt_format: str | None\n chosen_format: str | None\n rejected_format: str | None\n data_files: list[str] | None\n revision: str | None\n field_messages: str | None\n\n # For KTODataset:\n path: str | None\n split: str | None\n type: UserDefinedKTOType | str | None\n # For UserDefinedKTOType:\n field_system: str | None\n field_prompt: str | None\n field_completion: str | None\n field_label: bool | None\n prompt_format: str | None\n completion_format: str | None\n data_files: list[str] | None\n trust_remote_code: bool | None = False\n revision: str | None\n\n # For StepwiseSupervisedDataset:\n path: str | None\n split: str | None\n data_files: list[str] | None\n revision: str | None\n step_separator: str | None\n max_completion_length: int | None\n train_on_last_step_only: bool | None\n\n# A list of one or more datasets to eval the model with. You can use either\n# test_datasets, or val_set_size, but not both.\ntest_datasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None\n # For SFTDataset:\n # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory\n path: str | None\n # name of dataset split to load from\n split: str | None\n # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]\n type: str | UserDefinedPrompterType | None\n # For UserDefinedPrompterType:\n # Custom user instruction prompt\n system_prompt: str | None\n # Use {system} as key to be replaced\n system_format: str | None\n field_system: str | None\n field_instruction: str | None\n field_input: str | None\n field_output: str | None\n\n # Customizable to be single line or multi-line. Use {instruction}/{input} as key to\n # be replaced. 'format' can include {input}\n format: str | None\n # 'no_input_format' cannot include {input}\n no_input_format: str | None\n input_transform: str | None\n # split dataset into N pieces (use with shards_idx)\n shards: int | None\n # the index of sharded dataset to use\n shards_idx: int | None\n # process dataset in N sequential chunks for memory efficiency (exclusive with\n # `shards`)\n preprocess_shards: int | None\n conversation: str | None\n\n # The name of the chat template to use for training, following values are supported:\n # tokenizer_default: Uses the chat template that is available in the\n # tokenizer_config.json. If the chat template is not available in the tokenizer, it\n # will raise an error. This is the default.\n # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n # tokenizer_default_fallback_*: where * is the name of the chat template to fallback\n # to if the tokenizer does not have a chat template else default to tokenizer. E.g.\n # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat\n # template. The custom jinja template should be provided in the chat_template_jinja\n # field.\n chat_template: ChatTemplate | str | None\n # Custom jinja chat template or path to jinja file. Used only if `chat_template:\n # jinja` or empty.\n chat_template_jinja: str | None\n # path to source data files\n data_files: str | list[str] | None\n input_format: str | None\n # name of dataset configuration to load\n name: str | None\n # defines the datatype when path is a file\n ds_type: str | None\n # For `completion` datasets only, uses the provided field instead of `text` column\n field: str | None\n field_human: str | None\n field_model: str | None\n # Key containing the messages (default: \"messages\")\n field_messages: str | None\n # Key containing the tools (default: \"tools\"). Must be a list[dict] and follow [JSON\n # schema](https://json-schema.org/learn/getting-started-step-by-step).\n field_tools: str | None\n # Key containing the reasoning trace (default: \"reasoning_content\").\n field_thinking: str | None\n # The key the chat template expects that indicates the reasoning trace.\n template_thinking_key: str | None\n\n message_field_role: str | None\n\n message_field_content: str | None\n # Mapping of properties from the input dataset to the chat template. (default:\n # message_property_mappings={'role':'role', 'content':'content'}) If a property exists\n # in the template but not in this mapping, the system will attempt to load it directly\n # from the message using the property name as the key. Example: In the mapping below,\n # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and\n # used as 'content' in the chat template.\n message_property_mappings: dict[str, str] | None\n # The key in the message turn that indicates via boolean whether tokens of a turn\n # should be considered for training. Useful to selectively train on certain turns\n # besides the `roles_to_train`.\n message_field_training: str | None\n # The key in the message turn that contains the training details. Useful to\n # selectively train on certain tokens in a turn. The value of the key is a List[Dict]\n # containing `begin_offset` (start character index in content), `end_offset` (end\n # character index in content), and `train` (boolean whether to train).\n message_field_training_detail: str | None\n # (for Qwen3 template only) Whether to split the assistant content based on a\n # reasoning trace inside delimited tags\n split_thinking: bool | None\n logprobs_field: str | None\n temperature: float | None\n # Roles to train on. The tokens from these roles will be considered for the loss.\n roles_to_train: list[str] | None\n # Which EOS tokens to train on in the conversation. Possible values are: all: train on\n # all EOS tokens, turn (default): train on the EOS token at the end of each trainable\n # turn, last: train on the last EOS token in the conversation\n train_on_eos: Literal['all', 'turn', 'last'] | None\n # Roles mapping in the messages. The format is {target_role: [source_roles]}. All\n # source roles will be mapped to the target role. The default is: user: [\"human\",\n # \"user\"], assistant: [\"gpt\", \"assistant\"], system: [\"system\"], tool: [\"tool\"]\n roles: dict[str, list[str]] | None\n # Whether to drop the system turn from the dataset. Only works with chat_template.\n # This does not drop the default system message from chat_template if it exists. If\n # you wish to, we recommend using a custom jinja template with the default system\n # message removed or adding a system turn with empty content.\n drop_system_message: bool | None\n # Trust remote code for untrusted source\n trust_remote_code: bool | None = False\n # The specific revision of the dataset to use when loading from the Hugging Face Hub.\n # This can be a commit hash, tag, or branch name. If not specified, the latest version\n # will be used. This parameter is ignored for local datasets.\n revision: str | None\n\n # For DPODataset:\n path: str | None\n split: str | None\n type: UserDefinedDPOType | str | None\n # For UserDefinedDPOType:\n field_system: str | None\n field_prompt: str | None\n field_chosen: str | None\n field_rejected: str | None\n prompt_format: str | None\n chosen_format: str | None\n rejected_format: str | None\n data_files: list[str] | None\n revision: str | None\n field_messages: str | None\n\n # For KTODataset:\n path: str | None\n split: str | None\n type: UserDefinedKTOType | str | None\n # For UserDefinedKTOType:\n field_system: str | None\n field_prompt: str | None\n field_completion: str | None\n field_label: bool | None\n prompt_format: str | None\n completion_format: str | None\n data_files: list[str] | None\n trust_remote_code: bool | None = False\n revision: str | None\n\n # For StepwiseSupervisedDataset:\n path: str | None\n split: str | None\n data_files: list[str] | None\n revision: str | None\n step_separator: str | None\n max_completion_length: int | None\n train_on_last_step_only: bool | None\n\n# If false, the datasets will not be shuffled and will keep their original order in\n# `datasets`. The same applies to the `test_datasets` option and the\n# `pretraining_dataset` option. Default is true.\nshuffle_merged_datasets: bool | None = True\n# If true, each dataset in `datasets` will be shuffled before merging. This allows\n# curriculum learning strategies to be applied at the dataset level. Default is false.\nshuffle_before_merging_datasets: bool | None = False\n# Axolotl attempts to save the dataset as an arrow after packing the data together so\n# subsequent training attempts load faster, relative path\ndataset_prepared_path: str | None\n# Num shards for whole dataset\ndataset_shard_num: int | None\n# Index of shard to use for whole dataset\ndataset_shard_idx: int | None\nskip_prepare_dataset: bool | None = False\n# Number of shards to save the prepared dataset\nnum_dataset_shards_to_save: int | None\n\n# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize\npretraining_dataset: Annotated[list[PretrainingDataset | SFTDataset], MinLen(1)] | None\n # For PretrainingDataset:\n name: str | None\n path: str | None\n split: str | None = train\n text_column: str | None = text\n type: str | None = pretrain\n trust_remote_code: bool | None = False\n data_files: str | None\n skip: int | None\n\n # For SFTDataset:\n # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory\n path: str | None\n # name of dataset split to load from\n split: str | None\n # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]\n type: str | UserDefinedPrompterType | None\n # For UserDefinedPrompterType:\n # Custom user instruction prompt\n system_prompt: str | None\n # Use {system} as key to be replaced\n system_format: str | None\n field_system: str | None\n field_instruction: str | None\n field_input: str | None\n field_output: str | None\n\n # Customizable to be single line or multi-line. Use {instruction}/{input} as key to\n # be replaced. 'format' can include {input}\n format: str | None\n # 'no_input_format' cannot include {input}\n no_input_format: str | None\n input_transform: str | None\n # split dataset into N pieces (use with shards_idx)\n shards: int | None\n # the index of sharded dataset to use\n shards_idx: int | None\n # process dataset in N sequential chunks for memory efficiency (exclusive with\n # `shards`)\n preprocess_shards: int | None\n conversation: str | None\n\n # The name of the chat template to use for training, following values are supported:\n # tokenizer_default: Uses the chat template that is available in the\n # tokenizer_config.json. If the chat template is not available in the tokenizer, it\n # will raise an error. This is the default.\n # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n # tokenizer_default_fallback_*: where * is the name of the chat template to fallback\n # to if the tokenizer does not have a chat template else default to tokenizer. E.g.\n # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat\n # template. The custom jinja template should be provided in the chat_template_jinja\n # field.\n chat_template: ChatTemplate | str | None\n # Custom jinja chat template or path to jinja file. Used only if `chat_template:\n # jinja` or empty.\n chat_template_jinja: str | None\n # path to source data files\n data_files: str | list[str] | None\n input_format: str | None\n # name of dataset configuration to load\n name: str | None\n # defines the datatype when path is a file\n ds_type: str | None\n # For `completion` datasets only, uses the provided field instead of `text` column\n field: str | None\n field_human: str | None\n field_model: str | None\n # Key containing the messages (default: \"messages\")\n field_messages: str | None\n # Key containing the tools (default: \"tools\"). Must be a list[dict] and follow [JSON\n # schema](https://json-schema.org/learn/getting-started-step-by-step).\n field_tools: str | None\n # Key containing the reasoning trace (default: \"reasoning_content\").\n field_thinking: str | None\n # The key the chat template expects that indicates the reasoning trace.\n template_thinking_key: str | None\n\n message_field_role: str | None\n\n message_field_content: str | None\n # Mapping of properties from the input dataset to the chat template. (default:\n # message_property_mappings={'role':'role', 'content':'content'}) If a property exists\n # in the template but not in this mapping, the system will attempt to load it directly\n # from the message using the property name as the key. Example: In the mapping below,\n # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and\n # used as 'content' in the chat template.\n message_property_mappings: dict[str, str] | None\n # The key in the message turn that indicates via boolean whether tokens of a turn\n # should be considered for training. Useful to selectively train on certain turns\n # besides the `roles_to_train`.\n message_field_training: str | None\n # The key in the message turn that contains the training details. Useful to\n # selectively train on certain tokens in a turn. The value of the key is a List[Dict]\n # containing `begin_offset` (start character index in content), `end_offset` (end\n # character index in content), and `train` (boolean whether to train).\n message_field_training_detail: str | None\n # (for Qwen3 template only) Whether to split the assistant content based on a\n # reasoning trace inside delimited tags\n split_thinking: bool | None\n logprobs_field: str | None\n temperature: float | None\n # Roles to train on. The tokens from these roles will be considered for the loss.\n roles_to_train: list[str] | None\n # Which EOS tokens to train on in the conversation. Possible values are: all: train on\n # all EOS tokens, turn (default): train on the EOS token at the end of each trainable\n # turn, last: train on the last EOS token in the conversation\n train_on_eos: Literal['all', 'turn', 'last'] | None\n # Roles mapping in the messages. The format is {target_role: [source_roles]}. All\n # source roles will be mapped to the target role. The default is: user: [\"human\",\n # \"user\"], assistant: [\"gpt\", \"assistant\"], system: [\"system\"], tool: [\"tool\"]\n roles: dict[str, list[str]] | None\n # Whether to drop the system turn from the dataset. Only works with chat_template.\n # This does not drop the default system message from chat_template if it exists. If\n # you wish to, we recommend using a custom jinja template with the default system\n # message removed or adding a system turn with empty content.\n drop_system_message: bool | None\n # Trust remote code for untrusted source\n trust_remote_code: bool | None = False\n # The specific revision of the dataset to use when loading from the Hugging Face Hub.\n # This can be a commit hash, tag, or branch name. If not specified, the latest version\n # will be used. This parameter is ignored for local datasets.\n revision: str | None\n\n# The maximum number of processes to use while preprocessing your input dataset. This\n# defaults to `os.cpu_count()` if not set. For Runpod VMs, it will default to number of\n# vCPUs via RUNPOD_CPU_COUNT.\ndataset_processes: int | None\n# Deduplicates datasets and test_datasets with identical entries\ndataset_exact_deduplication: bool | None\n# Keep dataset in memory while preprocessing. Only needed if cached dataset is taking\n# too much storage\ndataset_keep_in_memory: bool | None\ndataloader_pin_memory: bool | None\ndataloader_num_workers: int | None\ndataloader_prefetch_factor: int | None\ndataloader_drop_last: bool | None\n\naccelerator_config: dict[str, Any] | None\n\nremove_unused_columns: bool | None\n\n# Push prepared dataset to hub - repo_org/repo_name\npush_dataset_to_hub: str | None\n# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private\n# datasets. Required to be true when used in combination with `push_dataset_to_hub`\nhf_use_auth_token: bool | None\n\ndevice: Any | None\n# Passed through to transformers when loading the model when launched without\n# accelerate. Use `sequential` when training w/ model parallelism to limit memory\ndevice_map: Any | None\nworld_size: int | None\n# Don't mess with this, it's here for accelerate and torchrun\nlocal_rank: int | None\nddp: bool | None\n\n# Seed for reproducibility\nseed: int | None\n# Advanced DDP Arguments - timeout\nddp_timeout: int | None\n# Advanced DDP Arguments - bucket cap in MB\nddp_bucket_cap_mb: int | None\n# Advanced DDP Arguments - broadcast buffers\nddp_broadcast_buffers: bool | None\nddp_find_unused_parameters: bool | None\n\n# Approximate number of predictions sent to wandb depending on batch size. Enabled above\n# 0. Default is 0\neval_table_size: int | None\n# Total number of tokens generated for predictions sent to wandb. Default is 128\neval_max_new_tokens: int | None\n# Whether to run causal language model evaluation for metrics in\n# `eval_causal_lm_metrics`\ndo_causal_lm_eval: bool | None\n# HF evaluate metrics used during evaluation. Default is ['sacrebleu', 'comet', 'ter',\n# 'chrf', 'perplexity']\neval_causal_lm_metrics: list[str] | None\ndo_bench_eval: bool | None\nbench_dataset: str | None\nbench_split: str | None\nmetric_for_best_model: str | None\ngreater_is_better: bool | None\n\n# High loss value, indicating the learning has broken down (a good estimate is ~2 times\n# the loss at the start of training)\nloss_watchdog_threshold: float | None\n# Number of high-loss steps in a row before the trainer aborts (default: 3)\nloss_watchdog_patience: int | None\n\n# Run garbage collection every `gc_steps` steps. -1 will run on epoch end and before\n# evaluations. Default is 0 (disabled).\ngc_steps: int | None\n\n# Use CUDA bf16. bool or 'full' for `bf16_full_eval`, or 'auto' for automatic detection.\n# require >=ampere\nbf16: Literal['auto'] | bool | None = auto\n# Use CUDA fp16\nfp16: bool | None\n# Enable FP8 mixed precision training using TorchAO. Best used in combination with\n# torch.compile.\nfp8: bool | None\n# Enable FSDP float8 all-gather optimization for FP8 training. Can improve training\n# speed by 10-15% when FSDP is enabled.\nfp8_enable_fsdp_float8_all_gather: bool | None\n# No AMP (automatic mixed precision) - require >=ampere\nbfloat16: bool | None\n# No AMP (automatic mixed precision)\nfloat16: bool | None\n# Use CUDA tf32 - require >=ampere\ntf32: bool | None\nfloat32: bool | None\n\n# Whether to use gradient checkpointing. Available options are: true, false, 'offload',\n# 'offload_disk'.\n# https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing\ngradient_checkpointing: Literal['offload', 'offload_disk'] | bool | None = False\n# Additional kwargs to pass to the trainer for gradient checkpointing\ngradient_checkpointing_kwargs: dict[str, Any] | None\n# Whether to offload activations. Available options are: true, false, 'legacy', 'disk'.\nactivation_offloading: Literal['legacy', 'disk'] | bool | None = False\n\nunfrozen_parameters: list[str] | None\n\n# The maximum length of an input to train with, this should typically be less than 2048\n# as most models have a token/context limit of 2048\nsequence_len: int = 512\n# What to do when a tokenized row exceeds sequence_len. 'drop' removes the row;\n# 'truncate' slices tensors to sequence_len. Defaults to 'drop' for backward\n# compatibility.\nexcess_length_strategy: Literal['drop', 'truncate'] | None\n# The maximum length of an input for evaluation. If not specified, defaults to\n# sequence_len\neval_sequence_len: int | None\nmin_sample_len: int | None\n# maximum prompt length for RL training\nmax_prompt_len: int = 512\n# Use efficient multi-packing with block diagonal attention and per sequence\n# position_ids. Recommend set to 'true'\nsample_packing: bool | None\n# The number of samples packed at a time. Increasing the following values helps with\n# packing, but usually only slightly (<%1.)\nsample_packing_group_size: int | None = 100000\n# The number of samples which can be packed into one sequence. Increase if using a large\n# sequence_len with many short samples.\nsample_packing_bin_size: int | None = 200\n# Whether to pack samples sequentially\nsample_packing_sequentially: bool | None\n# The multiprocessing start method to use for packing. Should be 'fork', 'spawn' or\n# 'forkserver'\nsample_packing_mp_start_method: str | None\n# Set to 'false' if getting errors during eval with sample_packing on\neval_sample_packing: bool | None\n# Pad inputs so each step uses constant sized buffers. This will reduce memory\n# fragmentation and may prevent OOMs, by re-using memory more efficiently. Defaults to\n# True if `sample_packing` enabled\npad_to_sequence_len: bool | None\n# Whether to use sequential sampling for curriculum learning\ncurriculum_sampling: bool | None\nmultipack_real_batches: bool | None\n\n# Use batch flattening for speedups when not using sample_packing\nbatch_flattening: Literal['auto'] | bool | None\n\nuse_pose: bool | None\npose_split_on_token_ids: list[int] | None\npose_max_context_len: int | None\npose_num_chunks: int | None\n\npretrain_multipack_buffer_size: int | None\n# whether to prevent cross attention for packed sequences during pretraining\npretrain_multipack_attn: bool | None = True\n# whether to concatenate samples during pretraining\npretraining_sample_concatenation: bool | None\n\n# Use streaming mode for loading datasets\nstreaming: bool | None\n# Buffer size for multipack streaming datasets\nstreaming_multipack_buffer_size: int | None = 10000\n\n# Whether to use xformers attention patch https://github.com/facebookresearch/xformers\nxformers_attention: bool | None\n# Whether to use scaled-dot-product attention https://pytorch.org/docs/stable/generated/\n# torch.nn.functional.scaled_dot_product_attention.html\nsdp_attention: bool | None\n# Shifted-sparse attention (only llama) - https://arxiv.org/pdf/2309.12307.pdf\ns2_attention: bool | None\nflex_attention: bool | None\nflex_attn_compile_kwargs: dict[str, Any] | None\n# Whether to use flash attention patch https://github.com/Dao-AILab/flash-attention\nflash_attention: bool | None\n# Whether to use flash-attention cross entropy implementation - advanced use only\nflash_attn_cross_entropy: bool | None\n# Whether to use flash-attention rms norm implementation - advanced use only\nflash_attn_rms_norm: bool | None\n# Whether to fuse part of the MLP into a single operation\nflash_attn_fuse_mlp: bool | None\n# Whether to use bettertransformers\nflash_optimum: bool | None\n\neager_attention: bool | None\n\n# Specify a custom attention implementation, used mostly for kernels.\nattn_implementation: str | None\n\nunsloth_cross_entropy_loss: bool | None\nunsloth_lora_mlp: bool | None\nunsloth_lora_qkv: bool | None\nunsloth_lora_o: bool | None\nunsloth_rms_norm: bool | None\nunsloth_rope: bool | None\n\n# Apply custom LoRA autograd functions and activation function Triton kernels for speed\n# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html\nlora_mlp_kernel: bool | None\n# Apply custom LoRA autograd functions and activation function Triton kernels for speed\n# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html\nlora_qkv_kernel: bool | None\n# Apply custom LoRA autograd functions and activation function Triton kernels for speed\n# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html\nlora_o_kernel: bool | None\n\n# Whether to use chunked cross entropy loss for memory efficiency\nchunked_cross_entropy: bool | None\n# Number of chunks to use for chunked cross entropy loss\nchunked_cross_entropy_num_chunks: int | None\n\n# Whether to use ALST tiled mlp for memory efficient long context\ntiled_mlp: bool | None\n\n# Number of shards to use for ALST tiled mlp. If unset, it will be set based on\n# seqlen/hidden_size\ntiled_mlp_num_shards: int | None\n\n# Whether to use original mlp for ALST tiled mlp. Otherwise uses a generic MLP based on\n# llama.\ntiled_mlp_use_original_mlp: bool | None = True\n\nllama4_linearized_experts: bool | None\n\n# Deepspeed config path. e.g., deepspeed_configs/zero3.json\ndeepspeed: str | dict[str, Any] | None\n# Whether to use deepcompile for faster training with deepspeed\ndeepcompile: bool | None\n# FSDP configuration\nfsdp: list[str] | None\n\n# FSDP configuration options\nfsdp_config: dict[str, Any] | None\n# FSDP version\nfsdp_version: int | None\nfsdp_final_state_dict_type: Literal['FULL_STATE_DICT', 'LOCAL_STATE_DICT', 'SHARDED_STATE_DICT'] | None\n\n# How much of the dataset to set aside as evaluation. 1 = 100%, 0.50 = 50%, etc. 0 for\n# no eval.\nval_set_size: float | None = 0.0\n\n# Number of devices to shard across. If not set, will use all available devices.\ndp_shard_size: int | None\n# Number of devices to replicate across.\ndp_replicate_size: int | None\n# Deprecated: use `context_parallel_size` instead\nsequence_parallel_degree: int | None\n# Set to a divisor of the number of GPUs available to split sequences into chunks of\n# equal size. Use in long context training to prevent OOM when sequences cannot fit into\n# a single GPU's VRAM. E.g., if 4 GPUs are available, set this value to 2 to split each\n# sequence into two equal-sized subsequences, or set to 4 to split into four equal-sized\n# subsequences. See https://docs.axolotl.ai/docs/sequence_parallelism.html for more\n# details.\ncontext_parallel_size: int | None\n# Optional; strides across the key dimension. Larger values use more memory but should\n# make training faster. Must evenly divide the number of KV heads in your model.\nheads_k_stride: int | None\n# One of 'varlen_llama3', 'batch_ring', 'batch_zigzag', 'batch_stripe'. Defaults to\n# 'varlen_llama3' in the sample packing case, and 'batch_ring' in the non-sample packing\n# case.\nring_attn_func: RingAttnFunc | None\n# Number of tensor parallel processes in TP group. Only supported with DeepSpeed AutoTP.\ntensor_parallel_size: int | None\n\n# Add or change special tokens. If you add tokens here, you don't need to add them to\n# the `tokens` list.\nspecial_tokens: SpecialTokensConfig | None\n # For SpecialTokensConfig:\n bos_token: str | None\n eos_token: str | None\n pad_token: str | None\n unk_token: str | None\n additional_special_tokens: list[str] | None\n\n# Add extra tokens to the tokenizer\ntokens: list[str] | None\n# Mapping token_id to new_token_string to override reserved added_tokens in the\n# tokenizer. Only works for tokens that are not part of the base vocab (aka are\n# added_tokens). Can be checked if they exist in tokenizer.json added_tokens.\nadded_tokens_overrides: dict[int, str] | None\n\n# Whether to use torch.compile and which backend to use. setting to `auto` will enable\n# torch compile when torch>=2.6.0\ntorch_compile: Literal['auto'] | bool | None\n# Backend to use for torch.compile\ntorch_compile_backend: str | None\ntorch_compile_mode: Literal['default', 'reduce-overhead', 'max-autotune'] | None\n\n# Maximum number of iterations to train for. It precedes num_epochs which means that if\n# both are set, num_epochs will not be guaranteed. e.g., when 1 epoch is 1000 steps =>\n# `num_epochs: 2` and `max_steps: 100` will train for 100 steps\nmax_steps: int | None\n# Number of warmup steps. Cannot use with warmup_ratio\nwarmup_steps: int | None\n# Warmup ratio. Cannot use with warmup_steps\nwarmup_ratio: float | None\n# Leave empty to eval at each epoch, integer for every N steps. float for fraction of\n# total steps\neval_steps: int | float | None\n# Number of times per epoch to run evals, mutually exclusive with eval_steps\nevals_per_epoch: int | None\n# Set to `no` to skip evaluation, `epoch` at end of each epoch, leave empty to infer\n# from `eval_steps`\neval_strategy: str | None\n\n# Leave empty to save at each epoch, integer for every N steps. float for fraction of\n# total steps\nsave_steps: int | float | None\n# Number of times per epoch to save a checkpoint, mutually exclusive with save_steps\nsaves_per_epoch: int | None\n# Set to `no` to skip checkpoint saves, `epoch` at end of each epoch, `best` when better\n# result is achieved, leave empty to infer from `save_steps`\nsave_strategy: str | None\n# Checkpoints saved at a time\nsave_total_limit: int | None\n# Whether to checkpoint a model after the first step of training. Defaults to False.\nsave_first_step: bool | None\n\n# Logging frequency\nlogging_steps: int | None\n# Stop training after this many evaluation losses have increased in a row. https://huggi\n# ngface.co/transformers/v4.2.2/_modules/transformers/trainer_callback.html#EarlyStoppin\n# gCallback\nearly_stopping_patience: int | None\nload_best_model_at_end: bool | None = False\n# Save only the model weights, skipping the optimizer. Using this means you can't resume\n# from checkpoints.\nsave_only_model: bool | None = False\n# Use tensorboard for logging\nuse_tensorboard: bool | None\n# Enable the pytorch profiler to capture the first N steps of training to the\n# output_dir. see https://pytorch.org/blog/understanding-gpu-memory-1/ for more\n# information. Snapshots can be visualized @ https://pytorch.org/memory_viz\nprofiler_steps: int | None\n# Which step to start the profiler at. Useful for only capturing a few steps mid-run.\nprofiler_steps_start: int | None = 0\n# bool of whether to report tokens per second at the end of training. This is not\n# supported with pre-training datasets.\ninclude_tokens_per_second: bool | None\n# bool of whether to report tokens per second per-gpu during training by measuring\n# throughput of non-padding tokens.\ninclude_tkps: bool | None = True\n# NEFT https://arxiv.org/abs/2310.05914, set this to a number (paper default is 5) to\n# add noise to embeddings. Currently only supported on Llama and Mistral\nneftune_noise_alpha: float | None\n\n# Parameter controlling the relative ratio loss weight in the ORPO loss. Passed to\n# `beta` in `ORPOConfig` due to trl mapping.\norpo_alpha: float | None\n# Weighting of NLL term in loss from RPO paper\nrpo_alpha: float | None\n# Target reward margin for the SimPO loss\nsimpo_gamma: float | None\n# Weight of the BC regularizer\ncpo_alpha: float | None\n\n# Factor for desirable loss term in KTO loss\nkto_desirable_weight: float | None\n# Factor for undesirable loss term in KTO loss\nkto_undesirable_weight: float | None\n# The beta parameter for the RL training\nrl_beta: float | None\n\n# Defines the max memory usage per gpu on the system. Passed through to transformers\n# when loading the model.\nmax_memory: dict[int | Literal['cpu', 'disk'], int | str] | None\n# Limit the memory for all available GPUs to this amount (if an integer, expressed in\n# gigabytes); default: unset\ngpu_memory_limit: int | str | None\n# Whether to use low_cpu_mem_usage\nlow_cpu_mem_usage: bool | None\n\n# The name of the chat template to use for training, following values are supported:\n# tokenizer_default: Uses the chat template that is available in the\n# tokenizer_config.json. If the chat template is not available in the tokenizer, it will\n# raise an error. This is the default value.\n# alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n# are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n# tokenizer_default_fallback_*: where * is the name of the chat template to fallback to.\n# E.g. tokenizer_default_fallback_chatml. This is useful when the chat template is not\n# available in the tokenizer. jinja: Uses a custom jinja template for the chat template.\n# The custom jinja template should be provided in the chat_template_jinja field. The\n# selected chat template will be saved to the tokenizer_config.json for easier\n# inferencing\nchat_template: ChatTemplate | Annotated[str, StringConstraints(pattern='^tokenizer_default_fallback_')] | None\n# Custom jinja template or path to jinja file for chat template. This will be only used\n# if chat_template is set to `jinja` or `null` (in which case chat_template is\n# automatically set to `jinja`). Default is null.\nchat_template_jinja: str | None\n# Additional kwargs to pass to the chat template. This is useful for customizing the\n# chat template. For example, you can pass `thinking=False` to add a generation prompt\n# to the chat template.\nchat_template_kwargs: dict[str, Any] | None\n# Custom EOT (End-of-Turn) tokens to mask/unmask during training. These tokens mark the\n# boundaries between conversation turns. For example: ['/INST', '</s>',\n# '[/SYSTEM_PROMPT]']. If not specified, defaults to just the model's eos_token. This is\n# useful for templates that use multiple delimiter tokens.\neot_tokens: list[str] | None\n# Changes the default system message. Currently only supports chatml.\ndefault_system_message: str | None\n\n# Token index or indices to adjust embedding weights to the mean of the other tokens.\n# This is useful when the model has untrained embeddings.\nfix_untrained_tokens: int | list[int] | None\n\nis_preprocess: bool | None\npreprocess_iterable: bool | None\n\n# Total number of tokens - internal use\ntotal_num_tokens: int | None\ntotal_supervised_tokens: int | None\n# You can set these packing optimizations AFTER starting a training at least once. The\n# trainer will provide recommended values for these values.\nsample_packing_eff_est: float | None\naxolotl_config_path: str | None\n\n# Internal use only - Used to identify which the model is based on\nis_falcon_derived_model: bool | None\n# Internal use only - Used to identify which the model is based on\nis_llama_derived_model: bool | None\n# Internal use only - Used to identify which the model is based on. Please note that if\n# you set this to true, `padding_side` will be set to 'left' by default\nis_mistral_derived_model: bool | None\n# Internal use only - Used to identify which the model is based on\nis_qwen_derived_model: bool | None\n\n# Add plugins to extend the pipeline. See `src/axolotl/integrations` for the available\n# plugins or doc below for more details.\n# https://docs.axolotl.ai/docs/custom_integrations.html\nplugins: list[str] | None\n\n# This is the huggingface model that contains *.pt, *.safetensors, or *.bin files. This\n# can also be a relative path to a model on disk\nbase_model: str (required)\n# If the base_model repo on hf hub doesn't include configuration .json files, You can\n# set that here, or leave this empty to default to base_model\nbase_model_config: str | None\ncls_model_config: str | None\n# Optional tokenizer configuration path in case you want to use a different tokenizer\n# than the one defined in the base model\ntokenizer_config: str | None\n# use_fast option for tokenizer loading from_pretrained, default to True\ntokenizer_use_fast: bool | None\n# Whether to use the legacy tokenizer setting, defaults to True\ntokenizer_legacy: bool | None\n# Whether to use mistral-common tokenizer. If set to True, it will use the mistral-\n# common tokenizer.\ntokenizer_use_mistral_common: bool | None\n# Corresponding tokenizer for the model AutoTokenizer is a good choice\ntokenizer_type: str | None\n# transformers processor class\nprocessor_type: str | None\n# Whether to save jinja files for tokenizer, transformers default is True\ntokenizer_save_jinja_files: bool | None = True\n# Trust remote code for untrusted source\ntrust_remote_code: bool | None\n\n# Don't move the model to the device before sharding. Set to `false` to revert to legacy\n# behavior.\nexperimental_skip_move_to_device: bool | None = True\n\n# Use custom kernels, e.g. MegaBlocks.\nuse_kernels: bool | None\n\n# Model loading quantization config\nmodel_quantization_config: Literal['Mxfp4Config'] | None\n# kwargs for model quantization config\nmodel_quantization_config_kwargs: dict[str, Any] | None\n\n# Where to save the full-finetuned model to\noutput_dir: str = ./model-out\n# push checkpoints to hub\nhub_model_id: str | None\n# how to push checkpoints to hub\nhub_strategy: str | None\n# Save model as safetensors (require safetensors package). Default True\nsave_safetensors: bool | None = True\n\n# This will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer\nload_in_8bit: bool | None = False\n# Use bitsandbytes 4 bit\nload_in_4bit: bool | None = False\n\n# If you want to use 'lora' or 'qlora' or leave blank to train all parameters in\n# original model\nadapter: str | None\n# If you already have a lora model trained that you want to load, put that here. This\n# means after training, if you want to test the model, you should set this to the value\n# of `output_dir`. Note that if you merge an adapter to the base model, a new\n# subdirectory `merged` will be created under the `output_dir`.\nlora_model_dir: str | None\nlora_r: int | None\nlora_alpha: int | None\nlora_fan_in_fan_out: bool | None\nlora_target_modules: str | list[str] | None\nlora_target_parameters: str | list[str] | None\n# If true, will target all linear modules\nlora_target_linear: bool | None\n# If you added new tokens to the tokenizer, you may need to save some LoRA modules\n# because they need to know the new tokens. For LLaMA and Mistral, you need to save\n# `embed_tokens` and `lm_head`. It may vary for other models. `embed_tokens` converts\n# tokens to embeddings, and `lm_head` converts embeddings to token probabilities.\nlora_modules_to_save: list[str] | None\nlora_dropout: float | None = 0.0\n# The layer indices to transform, otherwise, apply to all layers\npeft_layers_to_transform: list[int] | None\npeft_layers_pattern: list[str] | None\n\npeft: PeftConfig | None\n # For PeftConfig:\n # Configuration options for loftq initialization for LoRA\n loftq_config: LoftQConfig | None\n # For LoftQConfig:\n # typically 4 bits\n loftq_bits: int = 4\n\n# Whether to use DoRA.\npeft_use_dora: bool | None\n# Whether to use RSLoRA.\npeft_use_rslora: bool | None\n# List of layer indices to replicate.\npeft_layer_replication: list[tuple[int, int]] | None\n# How to initialize LoRA weights. Default to True which is MS original implementation.\npeft_init_lora_weights: bool | str | None\n# A list of token indices to fine-tune on the `embed_tokens` layer. Otherwise, a dict\n# mapping an embedding layer name to its trainable token indices. See\n# https://huggingface.co/docs/peft/v0.17.0/en/developer_guides/lora#efficiently-train-\n# tokens-alongside-lora\npeft_trainable_token_indices: list[int] | dict[str, list[int]] | None\n\n# load qlora model in sharded format for FSDP using answer.ai technique.\nqlora_sharded_model_loading: bool | None = False\n# Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it\n# takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge\nlora_on_cpu: bool | None\n# Whether you are training a 4-bit GPTQ quantized model\ngptq: bool | None\n# optional overrides to the bnb 4bit quantization configuration\nbnb_config_kwargs: dict[str, Any] | None\n\n# loraplus learning rate ratio lr_B / lr_A. Recommended value is 2^4.\nloraplus_lr_ratio: float | None\n# loraplus learning rate for lora embedding layers. Default value is 1e-6.\nloraplus_lr_embedding: float | None = 1e-06\n\nmerge_lora: bool | None\n\n# Whether to use ReLoRA. Use with jagged_restart_*steps options.\nrelora: bool | None\n# threshold for optimizer magnitude when pruning\nrelora_prune_ratio: float | None\n# True to perform lora weight merges on cpu during restarts, for modest gpu memory\n# savings\nrelora_cpu_offload: bool | None\n\n# how often to reset for jagged restarts\njagged_restart_steps: int | None\n# how many warmup steps to take after reset for jagged restarts\njagged_restart_warmup_steps: int | None\n# how many anneal steps to take before reset for jagged restarts\njagged_restart_anneal_steps: int | None\n\n# If greater than 1, backpropagation will be skipped and the gradients will be\n# accumulated for the given number of steps.\ngradient_accumulation_steps: int | None = 1\n# The number of samples to include in each batch. This is the number of samples sent to\n# each GPU. Batch size per gpu = micro_batch_size * gradient_accumulation_steps\nmicro_batch_size: int | None = 1\n# Total batch size, we do not recommended setting this manually\nbatch_size: int | None\n# per gpu micro batch size for evals, defaults to value of micro_batch_size\neval_batch_size: int | None\n\n# whether to find batch size that fits in memory. Passed to underlying transformers\n# Trainer\nauto_find_batch_size: bool | None\n\n# Whether to mask out or include the human's prompt from the training labels\ntrain_on_inputs: bool | None = False\n# Group similarly sized data to minimize padding. May be slower to start, as it must\n# download and sort the entire dataset. Note that training loss may have an oscillating\n# pattern with this enabled.\ngroup_by_length: bool | None\n\nlearning_rate: str | float (required)\nembedding_lr: float | None\nembedding_lr_scale: float | None\n# Specify weight decay\nweight_decay: float | None = 0.0\n# Specify optimizer\noptimizer: OptimizerNames | CustomSupportedOptimizers | None = OptimizerNames.ADAMW_TORCH_FUSED\n# Dictionary of arguments to pass to the optimizer\noptim_args: str | dict[str, Any] | None\n# The target modules to optimize, i.e. the module names that you would like to train,\n# right now this is used only for GaLore algorithm\noptim_target_modules: list[str] | Literal['all_linear'] | None\n# Path to torch distx for optim 'adamw_anyprecision'\ntorchdistx_path: str | None\nlr_scheduler: SchedulerType | Literal['one_cycle'] | Literal['rex'] | None = SchedulerType.COSINE\n# Specify a scheduler and kwargs to use with the optimizer\nlr_scheduler_kwargs: dict[str, Any] | None\nlr_quadratic_warmup: bool | None\n# decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of\n# peak lr\ncosine_min_lr_ratio: float | None\n# freeze lr at some percentage of the step, e.g. cosine_constant_lr_ratio=0.8 means\n# start cosine_min_lr at 80% of training step\ncosine_constant_lr_ratio: float | None\n# Learning rate div factor\nlr_div_factor: float | None\n\nlr_groups: list[LrGroup] | None\n # For LrGroup:\n name: str (required)\n modules: list[str] (required)\n lr: float (required)\n\n# adamw hyperparams\nadam_epsilon: float | None\n# only used for CAME Optimizer\nadam_epsilon2: float | None\n# adamw hyperparams\nadam_beta1: float | None\n# adamw hyperparams\nadam_beta2: float | None\n# only used for CAME Optimizer\nadam_beta3: float | None\n\n# Dion Optimizer learning rate\ndion_lr: float | None\n# Dion Optimizer momentum\ndion_momentum: float | None\n# Dion Optimizer: r/d fraction for low-rank approximation. Used to compute the low-rank\n# dimension.\ndion_rank_fraction: float | None = 1.0\n# Dion Optimizer: Round up the low-rank dimension to a multiple of this number. This may\n# be useful to ensure even sharding.\ndion_rank_multiple_of: int | None = 1\n\n# Gradient clipping max norm\nmax_grad_norm: float | None\nnum_epochs: float = 1.0\n\nuse_wandb: bool | None\n# Set the name of your wandb run\nwandb_name: str | None\n# Set the ID of your wandb run\nwandb_run_id: str | None\n# \"offline\" to save run metadata locally and not sync to the server, \"disabled\" to turn\n# off wandb\nwandb_mode: str | None\n# Your wandb project name\nwandb_project: str | None\n# A wandb Team name if using a Team\nwandb_entity: str | None\nwandb_watch: str | None\n# \"checkpoint\" to log model to wandb Artifacts every `save_steps` or \"end\" to log only\n# at the end of training\nwandb_log_model: str | None\n\nuse_mlflow: bool | None\n# URI to mlflow\nmlflow_tracking_uri: str | None\n# Your experiment name\nmlflow_experiment_name: str | None\n# Your run name\nmlflow_run_name: str | None\n# set to true to copy each saved checkpoint on each save to mlflow artifact registry\nhf_mlflow_log_artifacts: bool | None\n\n# Enable or disable Comet integration.\nuse_comet: bool | None\n# API key for Comet. Recommended to set via `comet login`.\ncomet_api_key: str | None\n# Workspace name in Comet. Defaults to the user's default workspace.\ncomet_workspace: str | None\n# Project name in Comet. Defaults to Uncategorized.\ncomet_project_name: str | None\n# Identifier for the experiment. Used to append data to an existing experiment or\n# control the key of new experiments. Default to a random key.\ncomet_experiment_key: str | None\n# Create a new experiment (\"create\") or log to an existing one (\"get\"). Default\n# (\"get_or_create\") auto-selects based on configuration.\ncomet_mode: str | None\n# Set to True to log data to Comet server, or False for offline storage. Default is\n# True.\ncomet_online: bool | None\n# Dictionary for additional configuration settings, see the doc for more details.\ncomet_experiment_config: dict[str, Any] | None\n\n# the number of activate layers in LISA\nlisa_n_layers: int | None\n# how often to switch layers in LISA\nlisa_step_interval: int | None\n# path under the model to access the layers\nlisa_layers_attribute: str | None = model.layers\n\ngradio_title: str | None\ngradio_share: bool | None\ngradio_server_name: str | None\ngradio_server_port: int | None\ngradio_max_new_tokens: int | None\ngradio_temperature: float | None\n\nuse_ray: bool = False\nray_run_name: str | None\nray_num_workers: int = 1\nresources_per_worker: dict\n\n# The size of the image to resize to. It can be an integer (resized into padded-square\n# image) or a tuple (width, height).If not provided, we will attempt to load from\n# preprocessor.size, otherwise, images won't be resized.\nimage_size: int | tuple[int, int] | None\n# The resampling algorithm to use for image resizing. Default is bilinear. Please refer\n# to PIL.Image.Resampling for more details.\nimage_resize_algorithm: Literal['bilinear', 'bicubic', 'lanczos'] | Resampling | None\n\n# optional overrides to the base model configuration\noverrides_of_model_config: dict[str, Any] | None\n# optional overrides the base model loading from_pretrained\noverrides_of_model_kwargs: dict[str, Any] | None\n# If you want to specify the type of model to load, AutoModelForCausalLM is a good\n# choice too\ntype_of_model: str | None\n# You can specify to choose a specific model revision from huggingface hub\nrevision_of_model: str | None\n\nmax_packed_sequence_len: int | None\nrope_scaling: Any | None\nnoisy_embedding_alpha: float | None\ndpo_beta: float | None\nevaluation_strategy: str | None",
+ "text": "# Allow overwrite yml config using from cli\nstrict: bool | None = False\n# Resume from a specific checkpoint dir\nresume_from_checkpoint: str | None\n# If resume_from_checkpoint isn't set and you simply want it to start where it left off.\n# Be careful with this being turned on between different models.\nauto_resume_from_checkpoints: bool | None\n# Resize the model embeddings when new tokens are added to multiples of 32. This is\n# reported to improve training speed on some models\nresize_token_embeddings_to_32x: bool | None\nmean_resizing_embeddings: bool | None = False\n\n# Whether to shrink the embeddings to len(tokenizer). By default, we won't shrink.\nshrink_embeddings: bool | None\n# Don't upcast the embeddings to float32 when using PEFT. Useful for low-VRAM GPUs\nembeddings_skip_upcast: bool | None\n# Reinitialize model weights randomly instead of loading pretrained weights\nreinit_weights: bool | None\n\n# module to custom trainer class to use for training\ntrainer_cls: str | None\n\n# Use RL training: 'dpo', 'ipo', 'kto', 'simpo', 'orpo', 'grpo'\nrl: RLType | None\n\ntrl: TRLConfig | None\n # For TRLConfig:\n # Beta parameter for the RL training. Same as `rl_beta`. Use\n beta: float | None\n # Maximum length of the completion for RL training.\n max_completion_length: int | None\n\n # Whether to use VLLM for RL training.\n use_vllm: bool = False\n # VLLM mode to use, one of 'server' or 'colocate'\n vllm_mode: Literal['server', 'colocate'] | None\n # Host of the vLLM server to connect to.\n vllm_server_host: str | None = 0.0.0.0\n # Port of the vLLM server to connect to.\n vllm_server_port: int | None = 8000\n # Total timeout (in seconds) to wait for the vLLM server to respond.\n vllm_server_timeout: int | None\n # Regex for vLLM guided decoding.\n vllm_guided_decoding_regex: str | None\n\n # List of reward functions to load. Paths must be importable from current dir.\n reward_funcs: list[str] | None\n # List of reward weights for the reward functions.\n reward_weights: list[float] | None\n # Number of generations to sample.\n num_generations: int | None\n # Whether to log completions.\n log_completions: bool | None = False\n # Number of completions to print when log_completions is True.\n num_completions_to_print: int | None\n # Controls whether importance sampling ratios are computed at the `'token'` or\n # `'sequence'` level. For GSPO, use `sequence`, default is None which corresponds to\n # the original GRPO paper.\n importance_sampling_level: Literal['sequence', 'token'] | None\n\n # Whether to sync the reference model.\n sync_ref_model: bool | None = False\n # Mixup alpha for the reference model.\n ref_model_mixup_alpha: float | None = 0.9\n # Sync steps for the reference model.\n ref_model_sync_steps: int | None = 64\n # Whether to scale rewards by their standard deviation.\n scale_rewards: bool = True\n\n # Sampling temperature for the GRPO policy.\n temperature: float | None\n # Top-p sampling probability for the generation policy.\n top_p: float | None\n # Top-k sampling for the generation policy.\n top_k: int | None\n # Minimum probability for the generation policy.\n min_p: float | None\n # Penalty for tokens that appear in prompt and generated text.\n repetition_penalty: float | None\n # Number of iterations per batch (μ) for GRPO.\n num_iterations: int | None\n # Epsilon value for clipping in the GRPO algorithm.\n epsilon: float | None\n # Upper-bound epsilon value for clipping in the GRPO algorithm.\n epsilon_high: float | None\n # Whether to use Liger loss for GRPO.\n use_liger_loss: bool | None\n # Loss formulation to use. Supported values: grpo, bnpo, dr_grpo.\n loss_type: str | None\n # Whether to exclude truncated completions from loss calculation.\n mask_truncated_completions: bool = False\n\nvllm: VllmConfig | None\n # For VllmConfig:\n # Device to use for VLLM\n device: str | None = auto\n # Tensor parallel size for VLLM\n tensor_parallel_size: int | None\n # Data parallel size for VLLM\n data_parallel_size: int | None\n # GPU memory utilization for VLLM\n gpu_memory_utilization: float | None = 0.9\n # Data type for VLLM\n dtype: str | None = auto\n # Maximum length of the model context for VLLM\n max_model_len: int | None\n # Enable prefix caching for VLLM\n enable_prefix_caching: bool | None\n # Host for the vLLM server to start on\n host: str | None = 0.0.0.0\n # Port of the vLLM server to start on\n port: int | None = 8000\n\n # Enable reasoning for VLLM\n enable_reasoning: bool | None\n # Reasoning parser for VLLM\n reasoning_parser: str | None\n\nqat: QATConfig | None\n # For QATConfig:\n # Fake quantization layout to use for activation quantization.\n activation_dtype: TorchAOQuantDType | None\n # Fake quantization layout to use for weight quantization.\n weight_dtype: TorchAOQuantDType = TorchAOQuantDType.int8\n # Quantize embedding\n quantize_embedding: bool | None = False\n # The number of elements in each group for per-group fake quantization\n group_size: int | None = 32\n # The number of steps to apply fake quantization after\n fake_quant_after_n_steps: int | None\n\nquantization: PTQConfig | None\n # For PTQConfig:\n # Fake quantization layout to use for weight quantization.\n weight_dtype: TorchAOQuantDType = TorchAOQuantDType.int8\n # Fake quantization layout to use for activation quantization.\n activation_dtype: TorchAOQuantDType | None\n # Whether to quantize the embedding layer.\n quantize_embedding: bool | None\n # The number of elements in each group for per-group fake quantization\n group_size: int | None = 32\n\n# Reward modelling: `True` or `False`\nreward_model: bool | None\n# Process reward modelling: `True` or `False`\nprocess_reward_model: bool | None\n# Coefficient to incentivize the reward model to output mean-zero rewards (proposed by\n# https://huggingface.co/papers/2312.09244, Eq. 2). Recommended value: `0.01`.\ncenter_rewards_coefficient: float | None\nnum_labels: int | None\n\n# Whether to perform weighting in DPO trainer\ndpo_use_weighting: bool | None\ndpo_use_logits_to_keep: bool | None\ndpo_label_smoothing: float | None\ndpo_norm_loss: bool | None\ndpo_padding_free: bool | None\ndpo_generate_during_eval: bool | None\n\n# A list of one or more datasets to finetune the model with\ndatasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None\n # For SFTDataset:\n # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory\n path: str | None\n # name of dataset split to load from\n split: str | None\n # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]\n type: str | UserDefinedPrompterType | None\n # For UserDefinedPrompterType:\n # Custom user instruction prompt\n system_prompt: str | None\n # Use {system} as key to be replaced\n system_format: str | None\n field_system: str | None\n field_instruction: str | None\n field_input: str | None\n field_output: str | None\n\n # Customizable to be single line or multi-line. Use {instruction}/{input} as key to\n # be replaced. 'format' can include {input}\n format: str | None\n # 'no_input_format' cannot include {input}\n no_input_format: str | None\n input_transform: str | None\n # split dataset into N pieces (use with shards_idx)\n shards: int | None\n # the index of sharded dataset to use\n shards_idx: int | None\n # process dataset in N sequential chunks for memory efficiency (exclusive with\n # `shards`)\n preprocess_shards: int | None\n conversation: str | None\n\n # The name of the chat template to use for training, following values are supported:\n # tokenizer_default: Uses the chat template that is available in the\n # tokenizer_config.json. If the chat template is not available in the tokenizer, it\n # will raise an error. This is the default.\n # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n # tokenizer_default_fallback_*: where * is the name of the chat template to fallback\n # to if the tokenizer does not have a chat template else default to tokenizer. E.g.\n # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat\n # template. The custom jinja template should be provided in the chat_template_jinja\n # field.\n chat_template: ChatTemplate | str | None\n # Custom jinja chat template or path to jinja file. Used only if `chat_template:\n # jinja` or empty.\n chat_template_jinja: str | None\n # path to source data files\n data_files: str | list[str] | None\n input_format: str | None\n # name of dataset configuration to load\n name: str | None\n # defines the datatype when path is a file\n ds_type: str | None\n # For `completion` datasets only, uses the provided field instead of `text` column\n field: str | None\n field_human: str | None\n field_model: str | None\n # Key containing the messages (default: \"messages\")\n field_messages: str | None\n # Key containing the tools (default: \"tools\"). Must be a list[dict] and follow [JSON\n # schema](https://json-schema.org/learn/getting-started-step-by-step).\n field_tools: str | None\n # Key containing the reasoning trace (default: \"reasoning_content\").\n field_thinking: str | None\n # The key the chat template expects that indicates the reasoning trace.\n template_thinking_key: str | None\n\n message_field_role: str | None\n\n message_field_content: str | None\n # Mapping of properties from the input dataset to the chat template. (default:\n # message_property_mappings={'role':'role', 'content':'content'}) If a property exists\n # in the template but not in this mapping, the system will attempt to load it directly\n # from the message using the property name as the key. Example: In the mapping below,\n # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and\n # used as 'content' in the chat template.\n message_property_mappings: dict[str, str] | None\n # The key in the message turn that indicates via boolean whether tokens of a turn\n # should be considered for training. Useful to selectively train on certain turns\n # besides the `roles_to_train`.\n message_field_training: str | None\n # The key in the message turn that contains the training details. Useful to\n # selectively train on certain tokens in a turn. The value of the key is a List[Dict]\n # containing `begin_offset` (start character index in content), `end_offset` (end\n # character index in content), and `train` (boolean whether to train).\n message_field_training_detail: str | None\n # (for Qwen3 template only) Whether to split the assistant content based on a\n # reasoning trace inside delimited tags\n split_thinking: bool | None\n logprobs_field: str | None\n temperature: float | None\n # Roles to train on. The tokens from these roles will be considered for the loss.\n roles_to_train: list[str] | None\n # Which EOS tokens to train on in the conversation. Possible values are: all: train on\n # all EOS tokens, turn (default): train on the EOS token at the end of each trainable\n # turn, last: train on the last EOS token in the conversation\n train_on_eos: Literal['all', 'turn', 'last'] | None\n # Roles mapping in the messages. The format is {target_role: [source_roles]}. All\n # source roles will be mapped to the target role. The default is: user: [\"human\",\n # \"user\"], assistant: [\"gpt\", \"assistant\"], system: [\"system\"], tool: [\"tool\"]\n roles: dict[str, list[str]] | None\n # Whether to drop the system turn from the dataset. Only works with chat_template.\n # This does not drop the default system message from chat_template if it exists. If\n # you wish to, we recommend using a custom jinja template with the default system\n # message removed or adding a system turn with empty content.\n drop_system_message: bool | None\n # Trust remote code for untrusted source\n trust_remote_code: bool | None = False\n # The specific revision of the dataset to use when loading from the Hugging Face Hub.\n # This can be a commit hash, tag, or branch name. If not specified, the latest version\n # will be used. This parameter is ignored for local datasets.\n revision: str | None\n\n # For DPODataset:\n path: str | None\n split: str | None\n type: UserDefinedDPOType | str | None\n # For UserDefinedDPOType:\n field_system: str | None\n field_prompt: str | None\n field_chosen: str | None\n field_rejected: str | None\n prompt_format: str | None\n chosen_format: str | None\n rejected_format: str | None\n data_files: list[str] | None\n revision: str | None\n field_messages: str | None\n\n # For KTODataset:\n path: str | None\n split: str | None\n type: UserDefinedKTOType | str | None\n # For UserDefinedKTOType:\n field_system: str | None\n field_prompt: str | None\n field_completion: str | None\n field_label: bool | None\n prompt_format: str | None\n completion_format: str | None\n data_files: list[str] | None\n trust_remote_code: bool | None = False\n revision: str | None\n\n # For StepwiseSupervisedDataset:\n path: str | None\n split: str | None\n data_files: list[str] | None\n revision: str | None\n step_separator: str | None\n max_completion_length: int | None\n train_on_last_step_only: bool | None\n\n# A list of one or more datasets to eval the model with. You can use either\n# test_datasets, or val_set_size, but not both.\ntest_datasets: Annotated[list[SFTDataset | DPODataset | KTODataset | StepwiseSupervisedDataset], MinLen(1)] | None\n # For SFTDataset:\n # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory\n path: str | None\n # name of dataset split to load from\n split: str | None\n # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]\n type: str | UserDefinedPrompterType | None\n # For UserDefinedPrompterType:\n # Custom user instruction prompt\n system_prompt: str | None\n # Use {system} as key to be replaced\n system_format: str | None\n field_system: str | None\n field_instruction: str | None\n field_input: str | None\n field_output: str | None\n\n # Customizable to be single line or multi-line. Use {instruction}/{input} as key to\n # be replaced. 'format' can include {input}\n format: str | None\n # 'no_input_format' cannot include {input}\n no_input_format: str | None\n input_transform: str | None\n # split dataset into N pieces (use with shards_idx)\n shards: int | None\n # the index of sharded dataset to use\n shards_idx: int | None\n # process dataset in N sequential chunks for memory efficiency (exclusive with\n # `shards`)\n preprocess_shards: int | None\n conversation: str | None\n\n # The name of the chat template to use for training, following values are supported:\n # tokenizer_default: Uses the chat template that is available in the\n # tokenizer_config.json. If the chat template is not available in the tokenizer, it\n # will raise an error. This is the default.\n # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n # tokenizer_default_fallback_*: where * is the name of the chat template to fallback\n # to if the tokenizer does not have a chat template else default to tokenizer. E.g.\n # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat\n # template. The custom jinja template should be provided in the chat_template_jinja\n # field.\n chat_template: ChatTemplate | str | None\n # Custom jinja chat template or path to jinja file. Used only if `chat_template:\n # jinja` or empty.\n chat_template_jinja: str | None\n # path to source data files\n data_files: str | list[str] | None\n input_format: str | None\n # name of dataset configuration to load\n name: str | None\n # defines the datatype when path is a file\n ds_type: str | None\n # For `completion` datasets only, uses the provided field instead of `text` column\n field: str | None\n field_human: str | None\n field_model: str | None\n # Key containing the messages (default: \"messages\")\n field_messages: str | None\n # Key containing the tools (default: \"tools\"). Must be a list[dict] and follow [JSON\n # schema](https://json-schema.org/learn/getting-started-step-by-step).\n field_tools: str | None\n # Key containing the reasoning trace (default: \"reasoning_content\").\n field_thinking: str | None\n # The key the chat template expects that indicates the reasoning trace.\n template_thinking_key: str | None\n\n message_field_role: str | None\n\n message_field_content: str | None\n # Mapping of properties from the input dataset to the chat template. (default:\n # message_property_mappings={'role':'role', 'content':'content'}) If a property exists\n # in the template but not in this mapping, the system will attempt to load it directly\n # from the message using the property name as the key. Example: In the mapping below,\n # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and\n # used as 'content' in the chat template.\n message_property_mappings: dict[str, str] | None\n # The key in the message turn that indicates via boolean whether tokens of a turn\n # should be considered for training. Useful to selectively train on certain turns\n # besides the `roles_to_train`.\n message_field_training: str | None\n # The key in the message turn that contains the training details. Useful to\n # selectively train on certain tokens in a turn. The value of the key is a List[Dict]\n # containing `begin_offset` (start character index in content), `end_offset` (end\n # character index in content), and `train` (boolean whether to train).\n message_field_training_detail: str | None\n # (for Qwen3 template only) Whether to split the assistant content based on a\n # reasoning trace inside delimited tags\n split_thinking: bool | None\n logprobs_field: str | None\n temperature: float | None\n # Roles to train on. The tokens from these roles will be considered for the loss.\n roles_to_train: list[str] | None\n # Which EOS tokens to train on in the conversation. Possible values are: all: train on\n # all EOS tokens, turn (default): train on the EOS token at the end of each trainable\n # turn, last: train on the last EOS token in the conversation\n train_on_eos: Literal['all', 'turn', 'last'] | None\n # Roles mapping in the messages. The format is {target_role: [source_roles]}. All\n # source roles will be mapped to the target role. The default is: user: [\"human\",\n # \"user\"], assistant: [\"gpt\", \"assistant\"], system: [\"system\"], tool: [\"tool\"]\n roles: dict[str, list[str]] | None\n # Whether to drop the system turn from the dataset. Only works with chat_template.\n # This does not drop the default system message from chat_template if it exists. If\n # you wish to, we recommend using a custom jinja template with the default system\n # message removed or adding a system turn with empty content.\n drop_system_message: bool | None\n # Trust remote code for untrusted source\n trust_remote_code: bool | None = False\n # The specific revision of the dataset to use when loading from the Hugging Face Hub.\n # This can be a commit hash, tag, or branch name. If not specified, the latest version\n # will be used. This parameter is ignored for local datasets.\n revision: str | None\n\n # For DPODataset:\n path: str | None\n split: str | None\n type: UserDefinedDPOType | str | None\n # For UserDefinedDPOType:\n field_system: str | None\n field_prompt: str | None\n field_chosen: str | None\n field_rejected: str | None\n prompt_format: str | None\n chosen_format: str | None\n rejected_format: str | None\n data_files: list[str] | None\n revision: str | None\n field_messages: str | None\n\n # For KTODataset:\n path: str | None\n split: str | None\n type: UserDefinedKTOType | str | None\n # For UserDefinedKTOType:\n field_system: str | None\n field_prompt: str | None\n field_completion: str | None\n field_label: bool | None\n prompt_format: str | None\n completion_format: str | None\n data_files: list[str] | None\n trust_remote_code: bool | None = False\n revision: str | None\n\n # For StepwiseSupervisedDataset:\n path: str | None\n split: str | None\n data_files: list[str] | None\n revision: str | None\n step_separator: str | None\n max_completion_length: int | None\n train_on_last_step_only: bool | None\n\n# If false, the datasets will not be shuffled and will keep their original order in\n# `datasets`. The same applies to the `test_datasets` option and the\n# `pretraining_dataset` option. Default is true.\nshuffle_merged_datasets: bool | None = True\n# If true, each dataset in `datasets` will be shuffled before merging. This allows\n# curriculum learning strategies to be applied at the dataset level. Default is false.\nshuffle_before_merging_datasets: bool | None = False\n# Axolotl attempts to save the dataset as an arrow after packing the data together so\n# subsequent training attempts load faster, relative path\ndataset_prepared_path: str | None\n# Num shards for whole dataset\ndataset_shard_num: int | None\n# Index of shard to use for whole dataset\ndataset_shard_idx: int | None\nskip_prepare_dataset: bool | None = False\n# Number of shards to save the prepared dataset\nnum_dataset_shards_to_save: int | None\n\n# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize\npretraining_dataset: Annotated[list[PretrainingDataset | SFTDataset], MinLen(1)] | None\n # For PretrainingDataset:\n name: str | None\n path: str | None\n split: str | None = train\n text_column: str | None = text\n type: str | None = pretrain\n trust_remote_code: bool | None = False\n data_files: str | None\n skip: int | None\n\n # For SFTDataset:\n # HuggingFace dataset repo | s3:// | gs:// | path to local file or directory\n path: str | None\n # name of dataset split to load from\n split: str | None\n # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]\n type: str | UserDefinedPrompterType | None\n # For UserDefinedPrompterType:\n # Custom user instruction prompt\n system_prompt: str | None\n # Use {system} as key to be replaced\n system_format: str | None\n field_system: str | None\n field_instruction: str | None\n field_input: str | None\n field_output: str | None\n\n # Customizable to be single line or multi-line. Use {instruction}/{input} as key to\n # be replaced. 'format' can include {input}\n format: str | None\n # 'no_input_format' cannot include {input}\n no_input_format: str | None\n input_transform: str | None\n # split dataset into N pieces (use with shards_idx)\n shards: int | None\n # the index of sharded dataset to use\n shards_idx: int | None\n # process dataset in N sequential chunks for memory efficiency (exclusive with\n # `shards`)\n preprocess_shards: int | None\n conversation: str | None\n\n # The name of the chat template to use for training, following values are supported:\n # tokenizer_default: Uses the chat template that is available in the\n # tokenizer_config.json. If the chat template is not available in the tokenizer, it\n # will raise an error. This is the default.\n # alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n # are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n # tokenizer_default_fallback_*: where * is the name of the chat template to fallback\n # to if the tokenizer does not have a chat template else default to tokenizer. E.g.\n # tokenizer_default_fallback_chatml. jinja: Uses a custom jinja template for the chat\n # template. The custom jinja template should be provided in the chat_template_jinja\n # field.\n chat_template: ChatTemplate | str | None\n # Custom jinja chat template or path to jinja file. Used only if `chat_template:\n # jinja` or empty.\n chat_template_jinja: str | None\n # path to source data files\n data_files: str | list[str] | None\n input_format: str | None\n # name of dataset configuration to load\n name: str | None\n # defines the datatype when path is a file\n ds_type: str | None\n # For `completion` datasets only, uses the provided field instead of `text` column\n field: str | None\n field_human: str | None\n field_model: str | None\n # Key containing the messages (default: \"messages\")\n field_messages: str | None\n # Key containing the tools (default: \"tools\"). Must be a list[dict] and follow [JSON\n # schema](https://json-schema.org/learn/getting-started-step-by-step).\n field_tools: str | None\n # Key containing the reasoning trace (default: \"reasoning_content\").\n field_thinking: str | None\n # The key the chat template expects that indicates the reasoning trace.\n template_thinking_key: str | None\n\n message_field_role: str | None\n\n message_field_content: str | None\n # Mapping of properties from the input dataset to the chat template. (default:\n # message_property_mappings={'role':'role', 'content':'content'}) If a property exists\n # in the template but not in this mapping, the system will attempt to load it directly\n # from the message using the property name as the key. Example: In the mapping below,\n # 'from' is loaded from input dataset and used as 'role', while 'value' is loaded and\n # used as 'content' in the chat template.\n message_property_mappings: dict[str, str] | None\n # The key in the message turn that indicates via boolean whether tokens of a turn\n # should be considered for training. Useful to selectively train on certain turns\n # besides the `roles_to_train`.\n message_field_training: str | None\n # The key in the message turn that contains the training details. Useful to\n # selectively train on certain tokens in a turn. The value of the key is a List[Dict]\n # containing `begin_offset` (start character index in content), `end_offset` (end\n # character index in content), and `train` (boolean whether to train).\n message_field_training_detail: str | None\n # (for Qwen3 template only) Whether to split the assistant content based on a\n # reasoning trace inside delimited tags\n split_thinking: bool | None\n logprobs_field: str | None\n temperature: float | None\n # Roles to train on. The tokens from these roles will be considered for the loss.\n roles_to_train: list[str] | None\n # Which EOS tokens to train on in the conversation. Possible values are: all: train on\n # all EOS tokens, turn (default): train on the EOS token at the end of each trainable\n # turn, last: train on the last EOS token in the conversation\n train_on_eos: Literal['all', 'turn', 'last'] | None\n # Roles mapping in the messages. The format is {target_role: [source_roles]}. All\n # source roles will be mapped to the target role. The default is: user: [\"human\",\n # \"user\"], assistant: [\"gpt\", \"assistant\"], system: [\"system\"], tool: [\"tool\"]\n roles: dict[str, list[str]] | None\n # Whether to drop the system turn from the dataset. Only works with chat_template.\n # This does not drop the default system message from chat_template if it exists. If\n # you wish to, we recommend using a custom jinja template with the default system\n # message removed or adding a system turn with empty content.\n drop_system_message: bool | None\n # Trust remote code for untrusted source\n trust_remote_code: bool | None = False\n # The specific revision of the dataset to use when loading from the Hugging Face Hub.\n # This can be a commit hash, tag, or branch name. If not specified, the latest version\n # will be used. This parameter is ignored for local datasets.\n revision: str | None\n\n# The maximum number of processes to use while preprocessing your input dataset. This\n# defaults to `os.cpu_count()` if not set. For Runpod VMs, it will default to number of\n# vCPUs via RUNPOD_CPU_COUNT.\ndataset_processes: int | None\n# Deduplicates datasets and test_datasets with identical entries\ndataset_exact_deduplication: bool | None\n# Keep dataset in memory while preprocessing. Only needed if cached dataset is taking\n# too much storage\ndataset_keep_in_memory: bool | None\ndataloader_pin_memory: bool | None\ndataloader_num_workers: int | None\ndataloader_prefetch_factor: int | None\ndataloader_drop_last: bool | None\n\naccelerator_config: dict[str, Any] | None\n\nremove_unused_columns: bool | None\n\n# Push prepared dataset to hub - repo_org/repo_name\npush_dataset_to_hub: str | None\n# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private\n# datasets. Required to be true when used in combination with `push_dataset_to_hub`\nhf_use_auth_token: bool | None\n\ndevice: Any | None\n# Passed through to transformers when loading the model when launched without\n# accelerate. Use `sequential` when training w/ model parallelism to limit memory\ndevice_map: Any | None\nworld_size: int | None\n# Don't mess with this, it's here for accelerate and torchrun\nlocal_rank: int | None\nddp: bool | None\n\n# Seed for reproducibility\nseed: int | None\n# Advanced DDP Arguments - timeout\nddp_timeout: int | None\n# Advanced DDP Arguments - bucket cap in MB\nddp_bucket_cap_mb: int | None\n# Advanced DDP Arguments - broadcast buffers\nddp_broadcast_buffers: bool | None\nddp_find_unused_parameters: bool | None\n\n# Approximate number of predictions sent to wandb depending on batch size. Enabled above\n# 0. Default is 0\neval_table_size: int | None\n# Total number of tokens generated for predictions sent to wandb. Default is 128\neval_max_new_tokens: int | None\n# Whether to run causal language model evaluation for metrics in\n# `eval_causal_lm_metrics`\ndo_causal_lm_eval: bool | None\n# HF evaluate metrics used during evaluation. Default is ['sacrebleu', 'comet', 'ter',\n# 'chrf', 'perplexity']\neval_causal_lm_metrics: list[str] | None\ndo_bench_eval: bool | None\nbench_dataset: str | None\nbench_split: str | None\nmetric_for_best_model: str | None\ngreater_is_better: bool | None\n\n# High loss value, indicating the learning has broken down (a good estimate is ~2 times\n# the loss at the start of training)\nloss_watchdog_threshold: float | None\n# Number of high-loss steps in a row before the trainer aborts (default: 3)\nloss_watchdog_patience: int | None\n\n# Run garbage collection every `gc_steps` steps. -1 will run on epoch end and before\n# evaluations. Default is 0 (disabled).\ngc_steps: int | None\n\n# Use CUDA bf16. bool or 'full' for `bf16_full_eval`, or 'auto' for automatic detection.\n# require >=ampere\nbf16: Literal['auto'] | bool | None = auto\n# Use CUDA fp16\nfp16: bool | None\n# Enable FP8 mixed precision training using TorchAO. Best used in combination with\n# torch.compile.\nfp8: bool | None\n# Enable FSDP float8 all-gather optimization for FP8 training. Can improve training\n# speed by 10-15% when FSDP is enabled.\nfp8_enable_fsdp_float8_all_gather: bool | None\n# No AMP (automatic mixed precision) - require >=ampere\nbfloat16: bool | None\n# No AMP (automatic mixed precision)\nfloat16: bool | None\n# Use CUDA tf32 - require >=ampere\ntf32: bool | None\nfloat32: bool | None\n\n# Whether to use gradient checkpointing. Available options are: true, false, 'offload',\n# 'offload_disk'.\n# https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing\ngradient_checkpointing: Literal['offload', 'offload_disk'] | bool | None = False\n# Additional kwargs to pass to the trainer for gradient checkpointing\ngradient_checkpointing_kwargs: dict[str, Any] | None\n# Whether to offload activations. Available options are: true, false, 'legacy', 'disk'.\nactivation_offloading: Literal['legacy', 'disk'] | bool | None = False\n\nunfrozen_parameters: list[str] | None\n\n# The maximum length of an input to train with, this should typically be less than 2048\n# as most models have a token/context limit of 2048\nsequence_len: int = 512\n# What to do when a tokenized row exceeds sequence_len. 'drop' removes the row;\n# 'truncate' slices tensors to sequence_len. Defaults to 'drop' for backward\n# compatibility.\nexcess_length_strategy: Literal['drop', 'truncate'] | None\n# The maximum length of an input for evaluation. If not specified, defaults to\n# sequence_len\neval_sequence_len: int | None\nmin_sample_len: int | None\n# maximum prompt length for RL training\nmax_prompt_len: int = 512\n# Use efficient multi-packing with block diagonal attention and per sequence\n# position_ids. Recommend set to 'true'\nsample_packing: bool | None\n# The number of samples packed at a time. Increasing the following values helps with\n# packing, but usually only slightly (<%1.)\nsample_packing_group_size: int | None = 100000\n# The number of samples which can be packed into one sequence. Increase if using a large\n# sequence_len with many short samples.\nsample_packing_bin_size: int | None = 200\n# Whether to pack samples sequentially\nsample_packing_sequentially: bool | None\n# The multiprocessing start method to use for packing. Should be 'fork', 'spawn' or\n# 'forkserver'\nsample_packing_mp_start_method: str | None\n# Set to 'false' if getting errors during eval with sample_packing on\neval_sample_packing: bool | None\n# Pad inputs so each step uses constant sized buffers. This will reduce memory\n# fragmentation and may prevent OOMs, by re-using memory more efficiently. Defaults to\n# True if `sample_packing` enabled\npad_to_sequence_len: bool | None\n# Whether to use sequential sampling for curriculum learning\ncurriculum_sampling: bool | None\nmultipack_real_batches: bool | None\n\n# Use batch flattening for speedups when not using sample_packing\nbatch_flattening: Literal['auto'] | bool | None\n\nuse_pose: bool | None\npose_split_on_token_ids: list[int] | None\npose_max_context_len: int | None\npose_num_chunks: int | None\n\npretrain_multipack_buffer_size: int | None\n# whether to prevent cross attention for packed sequences during pretraining\npretrain_multipack_attn: bool | None = True\n# whether to concatenate samples during pretraining\npretraining_sample_concatenation: bool | None\n\n# Use streaming mode for loading datasets\nstreaming: bool | None\n# Buffer size for multipack streaming datasets\nstreaming_multipack_buffer_size: int | None = 10000\n\n# Whether to use xformers attention patch https://github.com/facebookresearch/xformers\nxformers_attention: bool | None\n# Whether to use scaled-dot-product attention https://pytorch.org/docs/stable/generated/\n# torch.nn.functional.scaled_dot_product_attention.html\nsdp_attention: bool | None\n# Shifted-sparse attention (only llama) - https://arxiv.org/pdf/2309.12307.pdf\ns2_attention: bool | None\nflex_attention: bool | None\nflex_attn_compile_kwargs: dict[str, Any] | None\n# Whether to use flash attention patch https://github.com/Dao-AILab/flash-attention\nflash_attention: bool | None\n# Whether to use flash-attention cross entropy implementation - advanced use only\nflash_attn_cross_entropy: bool | None\n# Whether to use flash-attention rms norm implementation - advanced use only\nflash_attn_rms_norm: bool | None\n# Whether to fuse part of the MLP into a single operation\nflash_attn_fuse_mlp: bool | None\n# Whether to use bettertransformers\nflash_optimum: bool | None\n\neager_attention: bool | None\n\n# Specify a custom attention implementation, used mostly for kernels.\nattn_implementation: str | None\n\nunsloth_cross_entropy_loss: bool | None\nunsloth_lora_mlp: bool | None\nunsloth_lora_qkv: bool | None\nunsloth_lora_o: bool | None\nunsloth_rms_norm: bool | None\nunsloth_rope: bool | None\n\n# Apply custom LoRA autograd functions and activation function Triton kernels for speed\n# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html\nlora_mlp_kernel: bool | None\n# Apply custom LoRA autograd functions and activation function Triton kernels for speed\n# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html\nlora_qkv_kernel: bool | None\n# Apply custom LoRA autograd functions and activation function Triton kernels for speed\n# and memory savings. See: https://docs.axolotl.ai/docs/lora_optims.html\nlora_o_kernel: bool | None\n\n# Whether to use chunked cross entropy loss for memory efficiency\nchunked_cross_entropy: bool | None\n# Number of chunks to use for chunked cross entropy loss\nchunked_cross_entropy_num_chunks: int | None\n\n# Whether to use ALST tiled mlp for memory efficient long context\ntiled_mlp: bool | None\n\n# Number of shards to use for ALST tiled mlp. If unset, it will be set based on\n# seqlen/hidden_size\ntiled_mlp_num_shards: int | None\n\n# Whether to use original mlp for ALST tiled mlp. Otherwise uses a generic MLP based on\n# llama.\ntiled_mlp_use_original_mlp: bool | None = True\n\nllama4_linearized_experts: bool | None\n\n# Deepspeed config path. e.g., deepspeed_configs/zero3.json\ndeepspeed: str | dict[str, Any] | None\n# Whether to use deepcompile for faster training with deepspeed\ndeepcompile: bool | None\n# FSDP configuration\nfsdp: list[str] | None\n\n# FSDP configuration options\nfsdp_config: dict[str, Any] | None\n# FSDP version\nfsdp_version: int | None\nfsdp_final_state_dict_type: Literal['FULL_STATE_DICT', 'LOCAL_STATE_DICT', 'SHARDED_STATE_DICT'] | None\n\n# How much of the dataset to set aside as evaluation. 1 = 100%, 0.50 = 50%, etc. 0 for\n# no eval.\nval_set_size: float | None = 0.0\n\n# Number of devices to shard across. If not set, will use all available devices.\ndp_shard_size: int | None\n# Number of devices to replicate across.\ndp_replicate_size: int | None\n# Deprecated: use `context_parallel_size` instead\nsequence_parallel_degree: int | None\n# Set to a divisor of the number of GPUs available to split sequences into chunks of\n# equal size. Use in long context training to prevent OOM when sequences cannot fit into\n# a single GPU's VRAM. E.g., if 4 GPUs are available, set this value to 2 to split each\n# sequence into two equal-sized subsequences, or set to 4 to split into four equal-sized\n# subsequences. See https://docs.axolotl.ai/docs/sequence_parallelism.html for more\n# details.\ncontext_parallel_size: int | None\n# Optional; strides across the key dimension. Larger values use more memory but should\n# make training faster. Must evenly divide the number of KV heads in your model.\nheads_k_stride: int | None\n# One of 'varlen_llama3', 'batch_ring', 'batch_zigzag', 'batch_stripe'. Defaults to\n# 'varlen_llama3' in the sample packing case, and 'batch_ring' in the non-sample packing\n# case.\nring_attn_func: RingAttnFunc | None\n# Number of tensor parallel processes in TP group. Only supported with DeepSpeed AutoTP.\ntensor_parallel_size: int | None\n\n# Add or change special tokens. If you add tokens here, you don't need to add them to\n# the `tokens` list.\nspecial_tokens: SpecialTokensConfig | None\n # For SpecialTokensConfig:\n bos_token: str | None\n eos_token: str | None\n pad_token: str | None\n unk_token: str | None\n additional_special_tokens: list[str] | None\n\n# Add extra tokens to the tokenizer\ntokens: list[str] | None\n# Mapping token_id to new_token_string to override reserved added_tokens in the\n# tokenizer. Only works for tokens that are not part of the base vocab (aka are\n# added_tokens). Can be checked if they exist in tokenizer.json added_tokens.\nadded_tokens_overrides: dict[int, str] | None\n\n# Whether to use torch.compile and which backend to use. setting to `auto` will enable\n# torch compile when torch>=2.6.0\ntorch_compile: Literal['auto'] | bool | None\n# Backend to use for torch.compile\ntorch_compile_backend: str | None\ntorch_compile_mode: Literal['default', 'reduce-overhead', 'max-autotune'] | None\n\n# Maximum number of iterations to train for. It precedes num_epochs which means that if\n# both are set, num_epochs will not be guaranteed. e.g., when 1 epoch is 1000 steps =>\n# `num_epochs: 2` and `max_steps: 100` will train for 100 steps\nmax_steps: int | None\n# Number of warmup steps. Cannot use with warmup_ratio\nwarmup_steps: int | None\n# Warmup ratio. Cannot use with warmup_steps\nwarmup_ratio: float | None\n# Leave empty to eval at each epoch, integer for every N steps. float for fraction of\n# total steps\neval_steps: int | float | None\n# Number of times per epoch to run evals, mutually exclusive with eval_steps\nevals_per_epoch: int | None\n# Set to `no` to skip evaluation, `epoch` at end of each epoch, leave empty to infer\n# from `eval_steps`\neval_strategy: str | None\n\n# Leave empty to save at each epoch, integer for every N steps. float for fraction of\n# total steps\nsave_steps: int | float | None\n# Number of times per epoch to save a checkpoint, mutually exclusive with save_steps\nsaves_per_epoch: int | None\n# Set to `no` to skip checkpoint saves, `epoch` at end of each epoch, `best` when better\n# result is achieved, leave empty to infer from `save_steps`\nsave_strategy: str | None\n# Checkpoints saved at a time\nsave_total_limit: int | None\n# Whether to checkpoint a model after the first step of training. Defaults to False.\nsave_first_step: bool | None\n\n# Logging frequency\nlogging_steps: int | None\n# Stop training after this many evaluation losses have increased in a row. https://huggi\n# ngface.co/transformers/v4.2.2/_modules/transformers/trainer_callback.html#EarlyStoppin\n# gCallback\nearly_stopping_patience: int | None\nload_best_model_at_end: bool | None = False\n# Save only the model weights, skipping the optimizer. Using this means you can't resume\n# from checkpoints.\nsave_only_model: bool | None = False\n# Use tensorboard for logging\nuse_tensorboard: bool | None\n# Enable the pytorch profiler to capture the first N steps of training to the\n# output_dir. see https://pytorch.org/blog/understanding-gpu-memory-1/ for more\n# information. Snapshots can be visualized @ https://pytorch.org/memory_viz\nprofiler_steps: int | None\n# Which step to start the profiler at. Useful for only capturing a few steps mid-run.\nprofiler_steps_start: int | None = 0\n# bool of whether to report tokens per second at the end of training. This is not\n# supported with pre-training datasets.\ninclude_tokens_per_second: bool | None\n# bool of whether to report tokens per second per-gpu during training by measuring\n# throughput of non-padding tokens.\ninclude_tkps: bool | None = True\n# NEFT https://arxiv.org/abs/2310.05914, set this to a number (paper default is 5) to\n# add noise to embeddings. Currently only supported on Llama and Mistral\nneftune_noise_alpha: float | None\n\n# Parameter controlling the relative ratio loss weight in the ORPO loss. Passed to\n# `beta` in `ORPOConfig` due to trl mapping.\norpo_alpha: float | None\n# Weighting of NLL term in loss from RPO paper\nrpo_alpha: float | None\n# Target reward margin for the SimPO loss\nsimpo_gamma: float | None\n# Weight of the BC regularizer\ncpo_alpha: float | None\n\n# Factor for desirable loss term in KTO loss\nkto_desirable_weight: float | None\n# Factor for undesirable loss term in KTO loss\nkto_undesirable_weight: float | None\n# The beta parameter for the RL training\nrl_beta: float | None\n\n# Defines the max memory usage per gpu on the system. Passed through to transformers\n# when loading the model.\nmax_memory: dict[int | Literal['cpu', 'disk'], int | str] | None\n# Limit the memory for all available GPUs to this amount (if an integer, expressed in\n# gigabytes); default: unset\ngpu_memory_limit: int | str | None\n# Whether to use low_cpu_mem_usage\nlow_cpu_mem_usage: bool | None\n\n# The name of the chat template to use for training, following values are supported:\n# tokenizer_default: Uses the chat template that is available in the\n# tokenizer_config.json. If the chat template is not available in the tokenizer, it will\n# raise an error. This is the default value.\n# alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates\n# are available in the axolotl codebase at src/axolotl/utils/chat_templates.py.\n# tokenizer_default_fallback_*: where * is the name of the chat template to fallback to.\n# E.g. tokenizer_default_fallback_chatml. This is useful when the chat template is not\n# available in the tokenizer. jinja: Uses a custom jinja template for the chat template.\n# The custom jinja template should be provided in the chat_template_jinja field. The\n# selected chat template will be saved to the tokenizer_config.json for easier\n# inferencing\nchat_template: ChatTemplate | Annotated[str, StringConstraints(pattern='^tokenizer_default_fallback_')] | None\n# Custom jinja template or path to jinja file for chat template. This will be only used\n# if chat_template is set to `jinja` or `null` (in which case chat_template is\n# automatically set to `jinja`). Default is null.\nchat_template_jinja: str | None\n# Additional kwargs to pass to the chat template. This is useful for customizing the\n# chat template. For example, you can pass `thinking=False` to add a generation prompt\n# to the chat template.\nchat_template_kwargs: dict[str, Any] | None\n# Custom EOT (End-of-Turn) tokens to mask/unmask during training. These tokens mark the\n# boundaries between conversation turns. For example: ['/INST', '</s>',\n# '[/SYSTEM_PROMPT]']. If not specified, defaults to just the model's eos_token. This is\n# useful for templates that use multiple delimiter tokens.\neot_tokens: list[str] | None\n# Changes the default system message. Currently only supports chatml.\ndefault_system_message: str | None\n\n# Token index or indices to adjust embedding weights to the mean of the other tokens.\n# This is useful when the model has untrained embeddings.\nfix_untrained_tokens: int | list[int] | None\n\nis_preprocess: bool | None\npreprocess_iterable: bool | None\n\n# Total number of tokens - internal use\ntotal_num_tokens: int | None\ntotal_supervised_tokens: int | None\n# You can set these packing optimizations AFTER starting a training at least once. The\n# trainer will provide recommended values for these values.\nsample_packing_eff_est: float | None\naxolotl_config_path: str | None\n\n# Internal use only - Used to identify which the model is based on\nis_falcon_derived_model: bool | None\n# Internal use only - Used to identify which the model is based on\nis_llama_derived_model: bool | None\n# Internal use only - Used to identify which the model is based on. Please note that if\n# you set this to true, `padding_side` will be set to 'left' by default\nis_mistral_derived_model: bool | None\n# Internal use only - Used to identify which the model is based on\nis_qwen_derived_model: bool | None\n\n# Add plugins to extend the pipeline. See `src/axolotl/integrations` for the available\n# plugins or doc below for more details.\n# https://docs.axolotl.ai/docs/custom_integrations.html\nplugins: list[str] | None\n\n# This is the huggingface model that contains *.pt, *.safetensors, or *.bin files. This\n# can also be a relative path to a model on disk\nbase_model: str (required)\n# If the base_model repo on hf hub doesn't include configuration .json files, You can\n# set that here, or leave this empty to default to base_model\nbase_model_config: str | None\ncls_model_config: str | None\n# Optional tokenizer configuration path in case you want to use a different tokenizer\n# than the one defined in the base model\ntokenizer_config: str | None\n# use_fast option for tokenizer loading from_pretrained, default to True\ntokenizer_use_fast: bool | None\n# Whether to use the legacy tokenizer setting, defaults to True\ntokenizer_legacy: bool | None\n# Whether to use mistral-common tokenizer. If set to True, it will use the mistral-\n# common tokenizer.\ntokenizer_use_mistral_common: bool | None\n# Corresponding tokenizer for the model AutoTokenizer is a good choice\ntokenizer_type: str | None\n# transformers processor class\nprocessor_type: str | None\n# Whether to save jinja files for tokenizer, transformers default is True\ntokenizer_save_jinja_files: bool | None = True\n# Trust remote code for untrusted source\ntrust_remote_code: bool | None\n\n# Don't move the model to the device before sharding. Set to `false` to revert to legacy\n# behavior.\nexperimental_skip_move_to_device: bool | None = True\n\n# Use custom kernels, e.g. MegaBlocks.\nuse_kernels: bool | None\n\n# Model loading quantization config\nmodel_quantization_config: Literal['Mxfp4Config'] | None\n# kwargs for model quantization config\nmodel_quantization_config_kwargs: dict[str, Any] | None\n\n# Where to save the full-finetuned model to\noutput_dir: str = ./model-out\n# push checkpoints to hub\nhub_model_id: str | None\n# how to push checkpoints to hub\nhub_strategy: str | None\n# Save model as safetensors (require safetensors package). Default True\nsave_safetensors: bool | None = True\n\n# This will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer\nload_in_8bit: bool | None = False\n# Use bitsandbytes 4 bit\nload_in_4bit: bool | None = False\n\n# If you want to use 'lora' or 'qlora' or leave blank to train all parameters in\n# original model\nadapter: str | None\n# If you already have a lora model trained that you want to load, put that here. This\n# means after training, if you want to test the model, you should set this to the value\n# of `output_dir`. Note that if you merge an adapter to the base model, a new\n# subdirectory `merged` will be created under the `output_dir`.\nlora_model_dir: str | None\nlora_r: int | None\nlora_alpha: int | None\nlora_fan_in_fan_out: bool | None\nlora_target_modules: str | list[str] | None\nlora_target_parameters: str | list[str] | None\n# If true, will target all linear modules\nlora_target_linear: bool | None\n# If you added new tokens to the tokenizer, you may need to save some LoRA modules\n# because they need to know the new tokens. For LLaMA and Mistral, you need to save\n# `embed_tokens` and `lm_head`. It may vary for other models. `embed_tokens` converts\n# tokens to embeddings, and `lm_head` converts embeddings to token probabilities.\nlora_modules_to_save: list[str] | None\nlora_dropout: float | None = 0.0\n# The layer indices to transform, otherwise, apply to all layers\npeft_layers_to_transform: list[int] | None\npeft_layers_pattern: list[str] | None\n\npeft: PeftConfig | None\n # For PeftConfig:\n # Configuration options for loftq initialization for LoRA\n loftq_config: LoftQConfig | None\n # For LoftQConfig:\n # typically 4 bits\n loftq_bits: int = 4\n\n# Whether to use DoRA.\npeft_use_dora: bool | None\n# Whether to use RSLoRA.\npeft_use_rslora: bool | None\n# List of layer indices to replicate.\npeft_layer_replication: list[tuple[int, int]] | None\n# How to initialize LoRA weights. Default to True which is MS original implementation.\npeft_init_lora_weights: bool | str | None\n# A list of token indices to fine-tune on the `embed_tokens` layer. Otherwise, a dict\n# mapping an embedding layer name to its trainable token indices. See\n# https://huggingface.co/docs/peft/v0.17.0/en/developer_guides/lora#efficiently-train-\n# tokens-alongside-lora\npeft_trainable_token_indices: list[int] | dict[str, list[int]] | None\n\n# load qlora model in sharded format for FSDP using answer.ai technique.\nqlora_sharded_model_loading: bool | None = False\n# Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it\n# takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge\nlora_on_cpu: bool | None\n# Whether you are training a 4-bit GPTQ quantized model\ngptq: bool | None\n# optional overrides to the bnb 4bit quantization configuration\nbnb_config_kwargs: dict[str, Any] | None\n\n# loraplus learning rate ratio lr_B / lr_A. Recommended value is 2^4.\nloraplus_lr_ratio: float | None\n# loraplus learning rate for lora embedding layers. Default value is 1e-6.\nloraplus_lr_embedding: float | None = 1e-06\n\nmerge_lora: bool | None\n\n# Whether to use ReLoRA. Use with jagged_restart_*steps options.\nrelora: bool | None\n# threshold for optimizer magnitude when pruning\nrelora_prune_ratio: float | None\n# True to perform lora weight merges on cpu during restarts, for modest gpu memory\n# savings\nrelora_cpu_offload: bool | None\n\n# how often to reset for jagged restarts\njagged_restart_steps: int | None\n# how many warmup steps to take after reset for jagged restarts\njagged_restart_warmup_steps: int | None\n# how many anneal steps to take before reset for jagged restarts\njagged_restart_anneal_steps: int | None\n\n# If greater than 1, backpropagation will be skipped and the gradients will be\n# accumulated for the given number of steps.\ngradient_accumulation_steps: int | None = 1\n# The number of samples to include in each batch. This is the number of samples sent to\n# each GPU. Batch size per gpu = micro_batch_size * gradient_accumulation_steps\nmicro_batch_size: int | None = 1\n# Total batch size, we do not recommended setting this manually\nbatch_size: int | None\n# per gpu micro batch size for evals, defaults to value of micro_batch_size\neval_batch_size: int | None\n\n# whether to find batch size that fits in memory. Passed to underlying transformers\n# Trainer\nauto_find_batch_size: bool | None\n\n# Whether to mask out or include the human's prompt from the training labels\ntrain_on_inputs: bool | None = False\n# Group similarly sized data to minimize padding. May be slower to start, as it must\n# download and sort the entire dataset. Note that training loss may have an oscillating\n# pattern with this enabled.\ngroup_by_length: bool | None\n\nlearning_rate: str | float (required)\nembedding_lr: float | None\nembedding_lr_scale: float | None\n# Specify weight decay\nweight_decay: float | None = 0.0\n# Specify optimizer\noptimizer: OptimizerNames | CustomSupportedOptimizers | None = OptimizerNames.ADAMW_TORCH_FUSED\n# Dictionary of arguments to pass to the optimizer\noptim_args: str | dict[str, Any] | None\n# The target modules to optimize, i.e. the module names that you would like to train,\n# right now this is used only for GaLore algorithm\noptim_target_modules: list[str] | Literal['all_linear'] | None\n# Path to torch distx for optim 'adamw_anyprecision'\ntorchdistx_path: str | None\nlr_scheduler: SchedulerType | Literal['one_cycle'] | Literal['rex'] | None = SchedulerType.COSINE\n# Specify a scheduler and kwargs to use with the optimizer\nlr_scheduler_kwargs: dict[str, Any] | None\nlr_quadratic_warmup: bool | None\n# decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of\n# peak lr\ncosine_min_lr_ratio: float | None\n# freeze lr at some percentage of the step, e.g. cosine_constant_lr_ratio=0.8 means\n# start cosine_min_lr at 80% of training step\ncosine_constant_lr_ratio: float | None\n# Learning rate div factor\nlr_div_factor: float | None\n\nlr_groups: list[LrGroup] | None\n # For LrGroup:\n name: str (required)\n modules: list[str] (required)\n lr: float (required)\n\n# adamw hyperparams\nadam_epsilon: float | None\n# only used for CAME Optimizer\nadam_epsilon2: float | None\n# adamw hyperparams\nadam_beta1: float | None\n# adamw hyperparams\nadam_beta2: float | None\n# only used for CAME Optimizer\nadam_beta3: float | None\n\n# Dion Optimizer learning rate\ndion_lr: float | None\n# Dion Optimizer momentum\ndion_momentum: float | None\n# Dion Optimizer: r/d fraction for low-rank approximation. Used to compute the low-rank\n# dimension.\ndion_rank_fraction: float | None = 1.0\n# Dion Optimizer: Round up the low-rank dimension to a multiple of this number. This may\n# be useful to ensure even sharding.\ndion_rank_multiple_of: int | None = 1\n\n# Gradient clipping max norm\nmax_grad_norm: float | None\nnum_epochs: float = 1.0\n\nuse_wandb: bool | None\n# Set the name of your wandb run\nwandb_name: str | None\n# Set the ID of your wandb run\nwandb_run_id: str | None\n# \"offline\" to save run metadata locally and not sync to the server, \"disabled\" to turn\n# off wandb\nwandb_mode: str | None\n# Your wandb project name\nwandb_project: str | None\n# A wandb Team name if using a Team\nwandb_entity: str | None\nwandb_watch: str | None\n# \"checkpoint\" to log model to wandb Artifacts every `save_steps` or \"end\" to log only\n# at the end of training\nwandb_log_model: str | None\n\nuse_mlflow: bool | None\n# URI to mlflow\nmlflow_tracking_uri: str | None\n# Your experiment name\nmlflow_experiment_name: str | None\n# Your run name\nmlflow_run_name: str | None\n# set to true to copy each saved checkpoint on each save to mlflow artifact registry\nhf_mlflow_log_artifacts: bool | None\n\n# Enable or disable Comet integration.\nuse_comet: bool | None\n# API key for Comet. Recommended to set via `comet login`.\ncomet_api_key: str | None\n# Workspace name in Comet. Defaults to the user's default workspace.\ncomet_workspace: str | None\n# Project name in Comet. Defaults to Uncategorized.\ncomet_project_name: str | None\n# Identifier for the experiment. Used to append data to an existing experiment or\n# control the key of new experiments. Default to a random key.\ncomet_experiment_key: str | None\n# Create a new experiment (\"create\") or log to an existing one (\"get\"). Default\n# (\"get_or_create\") auto-selects based on configuration.\ncomet_mode: str | None\n# Set to True to log data to Comet server, or False for offline storage. Default is\n# True.\ncomet_online: bool | None\n# Dictionary for additional configuration settings, see the doc for more details.\ncomet_experiment_config: dict[str, Any] | None\n\n# the number of activate layers in LISA\nlisa_n_layers: int | None\n# how often to switch layers in LISA\nlisa_step_interval: int | None\n# path under the model to access the layers\nlisa_layers_attribute: str | None = model.layers\n\ngradio_title: str | None\ngradio_share: bool | None\ngradio_server_name: str | None\ngradio_server_port: int | None\ngradio_max_new_tokens: int | None\ngradio_temperature: float | None\n\nuse_ray: bool = False\nray_run_name: str | None\nray_num_workers: int = 1\nresources_per_worker: dict\n\n# The size of the image to resize to. It can be an integer (resized into padded-square\n# image) or a tuple (width, height).If not provided, we will attempt to load from\n# preprocessor.size, otherwise, images won't be resized.\nimage_size: int | tuple[int, int] | None\n# The resampling algorithm to use for image resizing. Default is bilinear. Please refer\n# to PIL.Image.Resampling for more details.\nimage_resize_algorithm: Literal['bilinear', 'bicubic', 'lanczos'] | Resampling | None\n\n# optional overrides to the base model configuration\noverrides_of_model_config: dict[str, Any] | None\n# optional overrides the base model loading from_pretrained\noverrides_of_model_kwargs: dict[str, Any] | None\n# If you want to specify the type of model to load, AutoModelForCausalLM is a good\n# choice too\ntype_of_model: str | None\n# You can specify to choose a specific model revision from huggingface hub\nrevision_of_model: str | None\n\nmax_packed_sequence_len: int | None\nrope_scaling: Any | None\nnoisy_embedding_alpha: float | None\ndpo_beta: float | None\nevaluation_strategy: str | None",
"crumbs": [
"Getting Started",
"Config Reference"
@@ -4051,6 +4051,16 @@
"section": "",
"text": "Acknowledgements\nPortions of this Cut Cross Entropy Software may utilize the following copyrighted\nmaterial, the use of which is hereby acknowledged.\n\nPyTorch\nFrom PyTorch:\n\nCopyright (c) 2016- Facebook, Inc (Adam Paszke)\nCopyright (c) 2014- Facebook, Inc (Soumith Chintala)\nCopyright (c) 2011-2014 Idiap Research Institute (Ronan Collobert)\nCopyright (c) 2012-2014 Deepmind Technologies (Koray Kavukcuoglu)\nCopyright (c) 2011-2012 NEC Laboratories America (Koray Kavukcuoglu)\nCopyright (c) 2011-2013 NYU (Clement Farabet)\nCopyright (c) 2006-2010 NEC Laboratories America (Ronan Collobert, Leon Bottou, Iain Melvin, Jason Weston)\nCopyright (c) 2006 Idiap Research Institute (Samy Bengio)\nCopyright (c) 2001-2004 Idiap Research Institute (Ronan Collobert, Samy Bengio, Johnny Mariethoz)\n\nFrom Caffe2:\n\nCopyright (c) 2016-present, Facebook Inc. All rights reserved.\n\nAll contributions by Facebook:\nCopyright (c) 2016 Facebook Inc.\n\nAll contributions by Google:\nCopyright (c) 2015 Google Inc.\nAll rights reserved.\n\nAll contributions by Yangqing Jia:\nCopyright (c) 2015 Yangqing Jia\nAll rights reserved.\n\nAll contributions by Kakao Brain:\nCopyright 2019-2020 Kakao Brain\n\nAll contributions by Cruise LLC:\nCopyright (c) 2022 Cruise LLC.\nAll rights reserved.\n\nAll contributions by Arm:\nCopyright (c) 2021, 2023-2024 Arm Limited and/or its affiliates\n\nAll contributions from Caffe:\nCopyright(c) 2013, 2014, 2015, the respective contributors\nAll rights reserved.\n\nAll other contributions:\nCopyright(c) 2015, 2016 the respective contributors\nAll rights reserved.\n\nCaffe2 uses a copyright model similar to Caffe: each contributor holds\ncopyright over their contributions to Caffe2. The project versioning records\nall such contribution and copyright details. If a contributor wants to further\nmark their specific copyright on a particular contribution, they should\nindicate their copyright solely in the commit message of the change when it is\ncommitted.\n\nAll rights reserved.\n\nRedistribution and use in source and binary forms, with or without\nmodification, are permitted provided that the following conditions are met:\n\n1. Redistributions of source code must retain the above copyright\nnotice, this list of conditions and the following disclaimer.\n\n2. Redistributions in binary form must reproduce the above copyright\nnotice, this list of conditions and the following disclaimer in the\ndocumentation and/or other materials provided with the distribution.\n\n3. Neither the names of Facebook, Deepmind Technologies, NYU, NEC Laboratories America\nand IDIAP Research Institute nor the names of its contributors may be\nused to endorse or promote products derived from this software without\nspecific prior written permission.\n\nTHIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\"\nAND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE\nIMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE\nARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE\nLIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR\nCONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF\nSUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS\nINTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN\nCONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)\nARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE\nPOSSIBILITY OF SUCH DAMAGE.\nTriton\n/*\n* Copyright 2018-2020 Philippe Tillet\n* Copyright 2020-2022 OpenAI\n*\n* Permission is hereby granted, free of charge, to any person obtaining\n* a copy of this software and associated documentation files\n* (the \"Software\"), to deal in the Software without restriction,\n* including without limitation the rights to use, copy, modify, merge,\n* publish, distribute, sublicense, and/or sell copies of the Software,\n* and to permit persons to whom the Software is furnished to do so,\n* subject to the following conditions:\n*\n* The above copyright notice and this permission notice shall be\n* included in all copies or substantial portions of the Software.\n*\n* THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND,\n* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF\n* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.\n* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY\n* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,\n* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE\n* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n*/\nTransformers\nCopyright 2018- The Hugging Face team. All rights reserved.\n\n Apache License\n Version 2.0, January 2004\n http://www.apache.org/licenses/\n\nTERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION\n\n1. Definitions.\n\n \"License\" shall mean the terms and conditions for use, reproduction,\n and distribution as defined by Sections 1 through 9 of this document.\n\n \"Licensor\" shall mean the copyright owner or entity authorized by\n the copyright owner that is granting the License.\n\n \"Legal Entity\" shall mean the union of the acting entity and all\n other entities that control, are controlled by, or are under common\n control with that entity. For the purposes of this definition,\n \"control\" means (i) the power, direct or indirect, to cause the\n direction or management of such entity, whether by contract or\n otherwise, or (ii) ownership of fifty percent (50%) or more of the\n outstanding shares, or (iii) beneficial ownership of such entity.\n\n \"You\" (or \"Your\") shall mean an individual or Legal Entity\n exercising permissions granted by this License.\n\n \"Source\" form shall mean the preferred form for making modifications,\n including but not limited to software source code, documentation\n source, and configuration files.\n\n \"Object\" form shall mean any form resulting from mechanical\n transformation or translation of a Source form, including but\n not limited to compiled object code, generated documentation,\n and conversions to other media types.\n\n \"Work\" shall mean the work of authorship, whether in Source or\n Object form, made available under the License, as indicated by a\n copyright notice that is included in or attached to the work\n (an example is provided in the Appendix below).\n\n \"Derivative Works\" shall mean any work, whether in Source or Object\n form, that is based on (or derived from) the Work and for which the\n editorial revisions, annotations, elaborations, or other modifications\n represent, as a whole, an original work of authorship. For the purposes\n of this License, Derivative Works shall not include works that remain\n separable from, or merely link (or bind by name) to the interfaces of,\n the Work and Derivative Works thereof.\n\n \"Contribution\" shall mean any work of authorship, including\n the original version of the Work and any modifications or additions\n to that Work or Derivative Works thereof, that is intentionally\n submitted to Licensor for inclusion in the Work by the copyright owner\n or by an individual or Legal Entity authorized to submit on behalf of\n the copyright owner. For the purposes of this definition, \"submitted\"\n means any form of electronic, verbal, or written communication sent\n to the Licensor or its representatives, including but not limited to\n communication on electronic mailing lists, source code control systems,\n and issue tracking systems that are managed by, or on behalf of, the\n Licensor for the purpose of discussing and improving the Work, but\n excluding communication that is conspicuously marked or otherwise\n designated in writing by the copyright owner as \"Not a Contribution.\"\n\n \"Contributor\" shall mean Licensor and any individual or Legal Entity\n on behalf of whom a Contribution has been received by Licensor and\n subsequently incorporated within the Work.\n\n2. Grant of Copyright License. Subject to the terms and conditions of\n this License, each Contributor hereby grants to You a perpetual,\n worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n copyright license to reproduce, prepare Derivative Works of,\n publicly display, publicly perform, sublicense, and distribute the\n Work and such Derivative Works in Source or Object form.\n\n3. Grant of Patent License. Subject to the terms and conditions of\n this License, each Contributor hereby grants to You a perpetual,\n worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n (except as stated in this section) patent license to make, have made,\n use, offer to sell, sell, import, and otherwise transfer the Work,\n where such license applies only to those patent claims licensable\n by such Contributor that are necessarily infringed by their\n Contribution(s) alone or by combination of their Contribution(s)\n with the Work to which such Contribution(s) was submitted. If You\n institute patent litigation against any entity (including a\n cross-claim or counterclaim in a lawsuit) alleging that the Work\n or a Contribution incorporated within the Work constitutes direct\n or contributory patent infringement, then any patent licenses\n granted to You under this License for that Work shall terminate\n as of the date such litigation is filed.\n\n4. Redistribution. You may reproduce and distribute copies of the\n Work or Derivative Works thereof in any medium, with or without\n modifications, and in Source or Object form, provided that You\n meet the following conditions:\n\n (a) You must give any other recipients of the Work or\n Derivative Works a copy of this License; and\n\n (b) You must cause any modified files to carry prominent notices\n stating that You changed the files; and\n\n (c) You must retain, in the Source form of any Derivative Works\n that You distribute, all copyright, patent, trademark, and\n attribution notices from the Source form of the Work,\n excluding those notices that do not pertain to any part of\n the Derivative Works; and\n\n (d) If the Work includes a \"NOTICE\" text file as part of its\n distribution, then any Derivative Works that You distribute must\n include a readable copy of the attribution notices contained\n within such NOTICE file, excluding those notices that do not\n pertain to any part of the Derivative Works, in at least one\n of the following places: within a NOTICE text file distributed\n as part of the Derivative Works; within the Source form or\n documentation, if provided along with the Derivative Works; or,\n within a display generated by the Derivative Works, if and\n wherever such third-party notices normally appear. The contents\n of the NOTICE file are for informational purposes only and\n do not modify the License. You may add Your own attribution\n notices within Derivative Works that You distribute, alongside\n or as an addendum to the NOTICE text from the Work, provided\n that such additional attribution notices cannot be construed\n as modifying the License.\n\n You may add Your own copyright statement to Your modifications and\n may provide additional or different license terms and conditions\n for use, reproduction, or distribution of Your modifications, or\n for any such Derivative Works as a whole, provided Your use,\n reproduction, and distribution of the Work otherwise complies with\n the conditions stated in this License.\n\n5. Submission of Contributions. Unless You explicitly state otherwise,\n any Contribution intentionally submitted for inclusion in the Work\n by You to the Licensor shall be under the terms and conditions of\n this License, without any additional terms or conditions.\n Notwithstanding the above, nothing herein shall supersede or modify\n the terms of any separate license agreement you may have executed\n with Licensor regarding such Contributions.\n\n6. Trademarks. This License does not grant permission to use the trade\n names, trademarks, service marks, or product names of the Licensor,\n except as required for reasonable and customary use in describing the\n origin of the Work and reproducing the content of the NOTICE file.\n\n7. Disclaimer of Warranty. Unless required by applicable law or\n agreed to in writing, Licensor provides the Work (and each\n Contributor provides its Contributions) on an \"AS IS\" BASIS,\n WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or\n implied, including, without limitation, any warranties or conditions\n of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A\n PARTICULAR PURPOSE. You are solely responsible for determining the\n appropriateness of using or redistributing the Work and assume any\n risks associated with Your exercise of permissions under this License.\n\n8. Limitation of Liability. In no event and under no legal theory,\n whether in tort (including negligence), contract, or otherwise,\n unless required by applicable law (such as deliberate and grossly\n negligent acts) or agreed to in writing, shall any Contributor be\n liable to You for damages, including any direct, indirect, special,\n incidental, or consequential damages of any character arising as a\n result of this License or out of the use or inability to use the\n Work (including but not limited to damages for loss of goodwill,\n work stoppage, computer failure or malfunction, or any and all\n other commercial damages or losses), even if such Contributor\n has been advised of the possibility of such damages.\n\n9. Accepting Warranty or Additional Liability. While redistributing\n the Work or Derivative Works thereof, You may choose to offer,\n and charge a fee for, acceptance of support, warranty, indemnity,\n or other liability obligations and/or rights consistent with this\n License. However, in accepting such obligations, You may act only\n on Your own behalf and on Your sole responsibility, not on behalf\n of any other Contributor, and only if You agree to indemnify,\n defend, and hold each Contributor harmless for any liability\n incurred by, or claims asserted against, such Contributor by reason\n of your accepting any such warranty or additional liability.\n\nEND OF TERMS AND CONDITIONS\n\nAPPENDIX: How to apply the Apache License to your work.\n\n To apply the Apache License to your work, attach the following\n boilerplate notice, with the fields enclosed by brackets \"[]\"\n replaced with your own identifying information. (Don't include\n the brackets!) The text should be enclosed in the appropriate\n comment syntax for the file format. We also recommend that a\n file or class name and description of purpose be included on the\n same \"printed page\" as the copyright notice for easier\n identification within third-party archives.\n\nCopyright [yyyy] [name of copyright owner]\n\nLicensed under the Apache License, Version 2.0 (the \"License\");\nyou may not use this file except in compliance with the License.\nYou may obtain a copy of the License at\n\n http://www.apache.org/licenses/LICENSE-2.0\n\nUnless required by applicable law or agreed to in writing, software\ndistributed under the License is distributed on an \"AS IS\" BASIS,\nWITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\nSee the License for the specific language governing permissions and\nlimitations under the License."
},
+ {
+ "objectID": "index.html",
+ "href": "index.html",
+ "title": "Axolotl",
+ "section": "",
+ "text": "A Free and Open Source LLM Fine-tuning Framework",
+ "crumbs": [
+ "Home"
+ ]
+ },
{
"objectID": "index.html#latest-updates",
"href": "index.html#latest-updates",
@@ -4066,17 +4076,17 @@
"href": "index.html#overview",
"title": "Axolotl",
"section": "✨ Overview",
- "text": "✨ Overview\nAxolotl is a tool designed to streamline post-training for various AI models.\nFeatures:\n\nMultiple Model Support: Train various models like LLaMA, Mistral, Mixtral, Pythia, and more. We are compatible with HuggingFace transformers causal language models.\nTraining Methods: Full fine-tuning, LoRA, QLoRA, GPTQ, QAT, Preference Tuning (DPO, IPO, KTO, ORPO), RL (GRPO), Multimodal, and Reward Modelling (RM) / Process Reward Modelling (PRM).\nEasy Configuration: Re-use a single YAML file between dataset preprocess, training, evaluation, quantization, and inference.\nPerformance Optimizations: Multipacking, Flash Attention, Xformers, Flex Attention, Liger Kernel, Cut Cross Entropy, Sequence Parallelism (SP), LoRA optimizations, Multi-GPU training (FSDP1, FSDP2, DeepSpeed), Multi-node training (Torchrun, Ray), and many more!\nFlexible Dataset Handling: Load from local, HuggingFace, and cloud (S3, Azure, GCP, OCI) datasets.\nCloud Ready: We ship Docker images and also PyPI packages for use on cloud platforms and local hardware.",
+ "text": "✨ Overview\nAxolotl is a free and open-source tool designed to streamline post-training and fine-tuning for the latest large language models (LLMs).\nFeatures:\n\nMultiple Model Support: Train various models like GPT-OSS, LLaMA, Mistral, Mixtral, Pythia, and many more models available on the Hugging Face Hub.\nMultimodal Training: Fine-tune vision-language models (VLMs) including LLaMA-Vision, Qwen2-VL, Pixtral, LLaVA, SmolVLM2, and audio models like Voxtral with image, video, and audio support.\nTraining Methods: Full fine-tuning, LoRA, QLoRA, GPTQ, QAT, Preference Tuning (DPO, IPO, KTO, ORPO), RL (GRPO), and Reward Modelling (RM) / Process Reward Modelling (PRM).\nEasy Configuration: Re-use a single YAML configuration file across the full fine-tuning pipeline: dataset preprocessing, training, evaluation, quantization, and inference.\nPerformance Optimizations: Multipacking, Flash Attention, Xformers, Flex Attention, Liger Kernel, Cut Cross Entropy, Sequence Parallelism (SP), LoRA optimizations, Multi-GPU training (FSDP1, FSDP2, DeepSpeed), Multi-node training (Torchrun, Ray), and many more!\nFlexible Dataset Handling: Load from local, HuggingFace, and cloud (S3, Azure, GCP, OCI) datasets.\nCloud Ready: We ship Docker images and also PyPI packages for use on cloud platforms and local hardware.",
"crumbs": [
"Home"
]
},
{
- "objectID": "index.html#quick-start",
- "href": "index.html#quick-start",
+ "objectID": "index.html#quick-start---llm-fine-tuning-in-minutes",
+ "href": "index.html#quick-start---llm-fine-tuning-in-minutes",
"title": "Axolotl",
- "section": "🚀 Quick Start",
- "text": "🚀 Quick Start\nRequirements:\n\nNVIDIA GPU (Ampere or newer for bf16 and Flash Attention) or AMD GPU\nPython 3.11\nPyTorch ≥2.6.0\n\n\nGoogle Colab\n\n\n\nOpen In Colab\n\n\n\n\nInstallation\n\nUsing pip\npip3 install -U packaging==23.2 setuptools==75.8.0 wheel ninja\npip3 install --no-build-isolation axolotl[flash-attn,deepspeed]\n\n# Download example axolotl configs, deepspeed configs\naxolotl fetch examples\naxolotl fetch deepspeed_configs # OPTIONAL\n\n\nUsing Docker\nInstalling with Docker can be less error prone than installing in your own environment.\ndocker run --gpus '\"all\"' --rm -it axolotlai/axolotl:main-latest\nOther installation approaches are described here.\n\n\nCloud Providers\n\n\nRunPod\nVast.ai\nPRIME Intellect\nModal\nNovita\nJarvisLabs.ai\nLatitude.sh\n\n\n\n\n\nYour First Fine-tune\n# Fetch axolotl examples\naxolotl fetch examples\n\n# Or, specify a custom path\naxolotl fetch examples --dest path/to/folder\n\n# Train a model using LoRA\naxolotl train examples/llama-3/lora-1b.yml\nThat’s it! Check out our Getting Started Guide for a more detailed walkthrough.",
+ "section": "🚀 Quick Start - LLM Fine-tuning in Minutes",
+ "text": "🚀 Quick Start - LLM Fine-tuning in Minutes\nRequirements:\n\nNVIDIA GPU (Ampere or newer for bf16 and Flash Attention) or AMD GPU\nPython 3.11\nPyTorch ≥2.6.0\n\n\nGoogle Colab\n\n\n\nOpen In Colab\n\n\n\n\nInstallation\n\nUsing pip\npip3 install -U packaging==23.2 setuptools==75.8.0 wheel ninja\npip3 install --no-build-isolation axolotl[flash-attn,deepspeed]\n\n# Download example axolotl configs, deepspeed configs\naxolotl fetch examples\naxolotl fetch deepspeed_configs # OPTIONAL\n\n\nUsing Docker\nInstalling with Docker can be less error prone than installing in your own environment.\ndocker run --gpus '\"all\"' --rm -it axolotlai/axolotl:main-latest\nOther installation approaches are described here.\n\n\nCloud Providers\n\n\nRunPod\nVast.ai\nPRIME Intellect\nModal\nNovita\nJarvisLabs.ai\nLatitude.sh\n\n\n\n\n\nYour First Fine-tune\n# Fetch axolotl examples\naxolotl fetch examples\n\n# Or, specify a custom path\naxolotl fetch examples --dest path/to/folder\n\n# Train a model using LoRA\naxolotl train examples/llama-3/lora-1b.yml\nThat’s it! Check out our Getting Started Guide for a more detailed walkthrough.",
"crumbs": [
"Home"
]
@@ -4126,7 +4136,7 @@
"href": "index.html#citing-axolotl",
"title": "Axolotl",
"section": "📝 Citing Axolotl",
- "text": "📝 Citing Axolotl\nIf you use Axolotl in your research or projects, please cite it as follows:\n@software{axolotl,\n title = {Axolotl: Post-Training for AI Models},\n author = {{Axolotl maintainers and contributors}},\n url = {https://github.com/axolotl-ai-cloud/axolotl},\n license = {Apache-2.0},\n year = {2023}\n}",
+ "text": "📝 Citing Axolotl\nIf you use Axolotl in your research or projects, please cite it as follows:\n@software{axolotl,\n title = {Axolotl: Open Source LLM Post-Training},\n author = {{Axolotl maintainers and contributors}},\n url = {https://github.com/axolotl-ai-cloud/axolotl},\n license = {Apache-2.0},\n year = {2023}\n}",
"crumbs": [
"Home"
]
diff --git a/sitemap.xml b/sitemap.xml
index ea3591c9d..ab466631a 100644
--- a/sitemap.xml
+++ b/sitemap.xml
@@ -2,794 +2,794 @@
https://docs.axolotl.ai/examples/colab-notebooks/colab-axolotl-example.html
- 2025-09-11T10:19:40.176Z
+ 2025-09-12T09:55:59.656Z
https://docs.axolotl.ai/src/axolotl/integrations/LICENSE.html
- 2025-09-11T10:19:40.193Z
+ 2025-09-12T09:55:59.673Z
https://docs.axolotl.ai/FAQS.html
- 2025-09-11T10:19:40.166Z
+ 2025-09-12T09:55:59.646Z
https://docs.axolotl.ai/docs/unsloth.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.653Z
https://docs.axolotl.ai/docs/reward_modelling.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/docker.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.649Z
https://docs.axolotl.ai/docs/batch_vs_grad.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/streaming.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/nccl.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/quantize.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/nd_parallelism.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/custom_integrations.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/debugging.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.649Z
https://docs.axolotl.ai/docs/gradient_checkpointing.html
- 2025-09-11T10:19:40.169Z
+ 2025-09-12T09:55:59.649Z
https://docs.axolotl.ai/docs/multimodal.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/dataset_loading.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/api/integrations.spectrum.args.html
- 2025-09-11T10:23:02.774Z
+ 2025-09-12T10:00:00.604Z
https://docs.axolotl.ai/docs/api/loaders.adapter.html
- 2025-09-11T10:23:01.897Z
+ 2025-09-12T09:59:59.753Z
https://docs.axolotl.ai/docs/api/core.builders.base.html
- 2025-09-11T10:23:01.533Z
+ 2025-09-12T09:59:59.394Z
https://docs.axolotl.ai/docs/api/cli.merge_sharded_fsdp_weights.html
- 2025-09-11T10:23:01.735Z
+ 2025-09-12T09:59:59.592Z
https://docs.axolotl.ai/docs/api/cli.cloud.modal_.html
- 2025-09-11T10:23:01.765Z
+ 2025-09-12T09:59:59.621Z
https://docs.axolotl.ai/docs/api/monkeypatch.stablelm_attn_hijack_flash.html
- 2025-09-11T10:23:02.315Z
+ 2025-09-12T10:00:00.169Z
https://docs.axolotl.ai/docs/api/integrations.kd.trainer.html
- 2025-09-11T10:23:02.764Z
+ 2025-09-12T10:00:00.594Z
https://docs.axolotl.ai/docs/api/cli.delinearize_llama4.html
- 2025-09-11T10:23:01.700Z
+ 2025-09-12T09:59:59.558Z
https://docs.axolotl.ai/docs/api/cli.evaluate.html
- 2025-09-11T10:23:01.648Z
+ 2025-09-12T09:59:59.506Z
https://docs.axolotl.ai/docs/api/utils.schemas.model.html
- 2025-09-11T10:23:02.529Z
+ 2025-09-12T10:00:00.376Z
https://docs.axolotl.ai/docs/api/cli.art.html
- 2025-09-11T10:23:01.671Z
+ 2025-09-12T09:59:59.529Z
https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.llama3.html
- 2025-09-11T10:23:02.083Z
+ 2025-09-12T09:59:59.935Z
https://docs.axolotl.ai/docs/api/core.chat.format.llama3x.html
- 2025-09-11T10:23:01.581Z
+ 2025-09-12T09:59:59.441Z
https://docs.axolotl.ai/docs/api/prompt_strategies.kto.chatml.html
- 2025-09-11T10:23:02.114Z
+ 2025-09-12T09:59:59.966Z
https://docs.axolotl.ai/docs/api/cli.args.html
- 2025-09-11T10:23:01.668Z
+ 2025-09-12T09:59:59.525Z
https://docs.axolotl.ai/docs/api/monkeypatch.gradient_checkpointing.offload_cpu.html
- 2025-09-11T10:23:02.340Z
+ 2025-09-12T10:00:00.194Z
https://docs.axolotl.ai/docs/api/evaluate.html
- 2025-09-11T10:23:01.456Z
+ 2025-09-12T09:59:59.316Z
https://docs.axolotl.ai/docs/api/prompt_strategies.input_output.html
- 2025-09-11T10:23:02.046Z
+ 2025-09-12T09:59:59.897Z
https://docs.axolotl.ai/docs/api/cli.vllm_serve.html
- 2025-09-11T10:23:01.755Z
+ 2025-09-12T09:59:59.611Z
https://docs.axolotl.ai/docs/api/cli.quantize.html
- 2025-09-11T10:23:01.748Z
+ 2025-09-12T09:59:59.605Z
https://docs.axolotl.ai/docs/api/cli.utils.sweeps.html
- 2025-09-11T10:23:01.794Z
+ 2025-09-12T09:59:59.651Z
https://docs.axolotl.ai/docs/api/core.builders.rl.html
- 2025-09-11T10:23:01.542Z
+ 2025-09-12T09:59:59.403Z
https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.chatml.html
- 2025-09-11T10:23:02.094Z
+ 2025-09-12T09:59:59.945Z
https://docs.axolotl.ai/docs/api/prompt_strategies.metharme.html
- 2025-09-11T10:23:02.057Z
+ 2025-09-12T09:59:59.908Z
https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.passthrough.html
- 2025-09-11T10:23:02.098Z
+ 2025-09-12T09:59:59.950Z
https://docs.axolotl.ai/docs/api/core.datasets.chat.html
- 2025-09-11T10:23:01.587Z
+ 2025-09-12T09:59:59.448Z
https://docs.axolotl.ai/docs/api/utils.distributed.html
- 2025-09-11T10:23:02.460Z
+ 2025-09-12T10:00:00.314Z
https://docs.axolotl.ai/docs/api/integrations.base.html
- 2025-09-11T10:23:02.752Z
+ 2025-09-12T10:00:00.582Z
https://docs.axolotl.ai/docs/api/index.html
- 2025-09-11T10:23:01.388Z
+ 2025-09-12T09:59:59.248Z
https://docs.axolotl.ai/docs/api/monkeypatch.gradient_checkpointing.offload_disk.html
- 2025-09-11T10:23:02.366Z
+ 2025-09-12T10:00:00.220Z
https://docs.axolotl.ai/docs/api/utils.quantization.html
- 2025-09-11T10:23:02.507Z
+ 2025-09-12T10:00:00.354Z
https://docs.axolotl.ai/docs/api/utils.dict.html
- 2025-09-11T10:23:02.465Z
+ 2025-09-12T10:00:00.319Z
https://docs.axolotl.ai/docs/api/core.trainers.dpo.trainer.html
- 2025-09-11T10:23:01.848Z
+ 2025-09-12T09:59:59.704Z
https://docs.axolotl.ai/docs/api/utils.schemas.config.html
- 2025-09-11T10:23:02.522Z
+ 2025-09-12T10:00:00.369Z
https://docs.axolotl.ai/docs/api/prompt_strategies.orcamini.html
- 2025-09-11T10:23:02.061Z
+ 2025-09-12T09:59:59.912Z
https://docs.axolotl.ai/docs/api/utils.collators.mamba.html
- 2025-09-11T10:23:02.816Z
+ 2025-09-12T10:00:00.647Z
https://docs.axolotl.ai/docs/api/datasets.html
- 2025-09-11T10:23:01.462Z
+ 2025-09-12T09:59:59.323Z
https://docs.axolotl.ai/docs/api/convert.html
- 2025-09-11T10:23:01.476Z
+ 2025-09-12T09:59:59.336Z
https://docs.axolotl.ai/docs/api/prompt_strategies.kto.llama3.html
- 2025-09-11T10:23:02.106Z
+ 2025-09-12T09:59:59.958Z
https://docs.axolotl.ai/docs/api/utils.tokenization.html
- 2025-09-11T10:23:02.372Z
+ 2025-09-12T10:00:00.226Z
https://docs.axolotl.ai/docs/api/core.trainers.base.html
- 2025-09-11T10:23:01.821Z
+ 2025-09-12T09:59:59.677Z
https://docs.axolotl.ai/docs/api/monkeypatch.llama_expand_mask.html
- 2025-09-11T10:23:02.270Z
+ 2025-09-12T10:00:00.124Z
https://docs.axolotl.ai/docs/api/loaders.model.html
- 2025-09-11T10:23:01.882Z
+ 2025-09-12T09:59:59.738Z
https://docs.axolotl.ai/docs/api/logging_config.html
- 2025-09-11T10:23:01.527Z
+ 2025-09-12T09:59:59.387Z
https://docs.axolotl.ai/docs/api/core.chat.format.shared.html
- 2025-09-11T10:23:01.582Z
+ 2025-09-12T09:59:59.443Z
https://docs.axolotl.ai/docs/api/models.mamba.modeling_mamba.html
- 2025-09-11T10:23:02.792Z
+ 2025-09-12T10:00:00.623Z
https://docs.axolotl.ai/docs/api/utils.schemas.enums.html
- 2025-09-11T10:23:02.593Z
+ 2025-09-12T10:00:00.439Z
https://docs.axolotl.ai/docs/api/utils.callbacks.profiler.html
- 2025-09-11T10:23:02.871Z
+ 2025-09-12T10:00:00.702Z
https://docs.axolotl.ai/docs/api/utils.bench.html
- 2025-09-11T10:23:02.388Z
+ 2025-09-12T10:00:00.242Z
https://docs.axolotl.ai/docs/api/utils.samplers.multipack.html
- 2025-09-11T10:23:02.861Z
+ 2025-09-12T10:00:00.692Z
https://docs.axolotl.ai/docs/api/core.trainers.mamba.html
- 2025-09-11T10:23:01.841Z
+ 2025-09-12T09:59:59.697Z
https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_chat.html
- 2025-09-11T10:23:02.000Z
+ 2025-09-12T09:59:59.851Z
https://docs.axolotl.ai/docs/api/monkeypatch.btlm_attn_hijack_flash.html
- 2025-09-11T10:23:02.308Z
+ 2025-09-12T10:00:00.162Z
https://docs.axolotl.ai/docs/api/core.chat.messages.html
- 2025-09-11T10:23:01.578Z
+ 2025-09-12T09:59:59.438Z
https://docs.axolotl.ai/docs/api/monkeypatch.llama_patch_multipack.html
- 2025-09-11T10:23:02.310Z
+ 2025-09-12T10:00:00.163Z
https://docs.axolotl.ai/docs/api/monkeypatch.mixtral.html
- 2025-09-11T10:23:02.337Z
+ 2025-09-12T10:00:00.191Z
https://docs.axolotl.ai/docs/api/core.trainers.mixins.scheduler.html
- 2025-09-11T10:23:01.928Z
+ 2025-09-12T09:59:59.780Z
https://docs.axolotl.ai/docs/api/loaders.patch_manager.html
- 2025-09-11T10:23:01.907Z
+ 2025-09-12T09:59:59.763Z
https://docs.axolotl.ai/docs/api/utils.optimizers.adopt.html
- 2025-09-11T10:23:02.473Z
+ 2025-09-12T10:00:00.327Z
https://docs.axolotl.ai/docs/api/monkeypatch.llama_attn_hijack_xformers.html
- 2025-09-11T10:23:02.263Z
+ 2025-09-12T10:00:00.116Z
https://docs.axolotl.ai/docs/api/utils.callbacks.perplexity.html
- 2025-09-11T10:23:02.867Z
+ 2025-09-12T10:00:00.698Z
https://docs.axolotl.ai/docs/api/loaders.tokenizer.html
- 2025-09-11T10:23:01.890Z
+ 2025-09-12T09:59:59.746Z
https://docs.axolotl.ai/docs/api/prompt_strategies.pygmalion.html
- 2025-09-11T10:23:02.067Z
+ 2025-09-12T09:59:59.919Z
https://docs.axolotl.ai/docs/api/utils.lora.html
- 2025-09-11T10:23:02.379Z
+ 2025-09-12T10:00:00.233Z
https://docs.axolotl.ai/docs/api/prompt_strategies.user_defined.html
- 2025-09-11T10:23:02.021Z
+ 2025-09-12T09:59:59.873Z
https://docs.axolotl.ai/docs/api/core.trainers.mixins.rng_state_loader.html
- 2025-09-11T10:23:01.922Z
+ 2025-09-12T09:59:59.773Z
https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.user_defined.html
- 2025-09-11T10:23:02.097Z
+ 2025-09-12T09:59:59.948Z
https://docs.axolotl.ai/docs/api/kernels.swiglu.html
- 2025-09-11T10:23:02.247Z
+ 2025-09-12T10:00:00.101Z
https://docs.axolotl.ai/docs/api/utils.schemas.multimodal.html
- 2025-09-11T10:23:02.571Z
+ 2025-09-12T10:00:00.418Z
https://docs.axolotl.ai/docs/api/cli.main.html
- 2025-09-11T10:23:01.632Z
+ 2025-09-12T09:59:59.489Z
https://docs.axolotl.ai/docs/api/utils.ctx_managers.sequence_parallel.html
- 2025-09-11T10:23:01.952Z
+ 2025-09-12T09:59:59.803Z
https://docs.axolotl.ai/docs/api/monkeypatch.multipack.html
- 2025-09-11T10:23:02.265Z
+ 2025-09-12T10:00:00.119Z
https://docs.axolotl.ai/docs/api/core.chat.format.chatml.html
- 2025-09-11T10:23:01.579Z
+ 2025-09-12T09:59:59.440Z
https://docs.axolotl.ai/docs/api/cli.cloud.base.html
- 2025-09-11T10:23:01.758Z
+ 2025-09-12T09:59:59.615Z
https://docs.axolotl.ai/docs/api/core.trainers.trl.html
- 2025-09-11T10:23:01.836Z
+ 2025-09-12T09:59:59.692Z
https://docs.axolotl.ai/docs/mixed_precision.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/installation.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.651Z
https://docs.axolotl.ai/docs/mac.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/sequence_parallelism.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/faq.html
- 2025-09-11T10:19:40.169Z
+ 2025-09-12T09:55:59.649Z
https://docs.axolotl.ai/docs/dataset-formats/tokenized.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/dataset-formats/stepwise_supervised.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/dataset-formats/pretraining.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/dataset-formats/template_free.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/dataset-formats/index.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/dataset-formats/conversation.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/dataset-formats/inst_tune.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/lr_groups.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/inference.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.651Z
https://docs.axolotl.ai/docs/lora_optims.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.651Z
https://docs.axolotl.ai/docs/multipack.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/amd_hpc.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/api/monkeypatch.trainer_fsdp_optim.html
- 2025-09-11T10:23:02.318Z
+ 2025-09-12T10:00:00.172Z
https://docs.axolotl.ai/docs/api/cli.inference.html
- 2025-09-11T10:23:01.715Z
+ 2025-09-12T09:59:59.572Z
https://docs.axolotl.ai/docs/api/utils.collators.mm_chat.html
- 2025-09-11T10:23:02.821Z
+ 2025-09-12T10:00:00.651Z
https://docs.axolotl.ai/docs/api/utils.model_shard_quant.html
- 2025-09-11T10:23:02.384Z
+ 2025-09-12T10:00:00.238Z
https://docs.axolotl.ai/docs/api/monkeypatch.llama_attn_hijack_flash.html
- 2025-09-11T10:23:02.261Z
+ 2025-09-12T10:00:00.115Z
https://docs.axolotl.ai/docs/api/integrations.grokfast.optimizer.html
- 2025-09-11T10:23:02.756Z
+ 2025-09-12T10:00:00.587Z
https://docs.axolotl.ai/docs/api/utils.schemas.training.html
- 2025-09-11T10:23:02.536Z
+ 2025-09-12T10:00:00.383Z
https://docs.axolotl.ai/docs/api/utils.schemas.integrations.html
- 2025-09-11T10:23:02.583Z
+ 2025-09-12T10:00:00.431Z
https://docs.axolotl.ai/docs/api/utils.schemas.utils.html
- 2025-09-11T10:23:02.599Z
+ 2025-09-12T10:00:00.445Z
https://docs.axolotl.ai/docs/api/cli.checks.html
- 2025-09-11T10:23:01.678Z
+ 2025-09-12T09:59:59.535Z
https://docs.axolotl.ai/docs/api/utils.callbacks.qat.html
- 2025-09-11T10:23:02.886Z
+ 2025-09-12T10:00:00.717Z
https://docs.axolotl.ai/docs/api/monkeypatch.mistral_attn_hijack_flash.html
- 2025-09-11T10:23:02.264Z
+ 2025-09-12T10:00:00.118Z
https://docs.axolotl.ai/docs/api/utils.data.sft.html
- 2025-09-11T10:23:02.482Z
+ 2025-09-12T10:00:00.335Z
https://docs.axolotl.ai/docs/api/utils.chat_templates.html
- 2025-09-11T10:23:02.374Z
+ 2025-09-12T10:00:00.228Z
https://docs.axolotl.ai/docs/api/monkeypatch.utils.html
- 2025-09-11T10:23:02.307Z
+ 2025-09-12T10:00:00.160Z
https://docs.axolotl.ai/docs/api/cli.utils.fetch.html
- 2025-09-11T10:23:01.783Z
+ 2025-09-12T09:59:59.639Z
https://docs.axolotl.ai/docs/api/core.builders.causal.html
- 2025-09-11T10:23:01.538Z
+ 2025-09-12T09:59:59.398Z
https://docs.axolotl.ai/docs/api/utils.schemas.trl.html
- 2025-09-11T10:23:02.566Z
+ 2025-09-12T10:00:00.413Z
https://docs.axolotl.ai/docs/api/utils.data.streaming.html
- 2025-09-11T10:23:02.474Z
+ 2025-09-12T10:00:00.328Z
https://docs.axolotl.ai/docs/api/monkeypatch.unsloth_.html
- 2025-09-11T10:23:02.326Z
+ 2025-09-12T10:00:00.180Z
https://docs.axolotl.ai/docs/api/prompt_strategies.base.html
- 2025-09-11T10:23:01.953Z
+ 2025-09-12T09:59:59.805Z
https://docs.axolotl.ai/docs/api/cli.merge_lora.html
- 2025-09-11T10:23:01.723Z
+ 2025-09-12T09:59:59.580Z
https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_instruct.html
- 2025-09-11T10:23:02.001Z
+ 2025-09-12T09:59:59.853Z
https://docs.axolotl.ai/docs/api/loaders.constants.html
- 2025-09-11T10:23:01.909Z
+ 2025-09-12T09:59:59.764Z
https://docs.axolotl.ai/docs/api/monkeypatch.transformers_fa_utils.html
- 2025-09-11T10:23:02.325Z
+ 2025-09-12T10:00:00.179Z
https://docs.axolotl.ai/docs/api/monkeypatch.relora.html
- 2025-09-11T10:23:02.269Z
+ 2025-09-12T10:00:00.123Z
https://docs.axolotl.ai/docs/api/utils.schedulers.html
- 2025-09-11T10:23:02.440Z
+ 2025-09-12T10:00:00.294Z
https://docs.axolotl.ai/docs/api/kernels.lora.html
- 2025-09-11T10:23:02.226Z
+ 2025-09-12T10:00:00.079Z
https://docs.axolotl.ai/docs/api/utils.collators.batching.html
- 2025-09-11T10:23:02.813Z
+ 2025-09-12T10:00:00.643Z
https://docs.axolotl.ai/docs/api/prompt_strategies.completion.html
- 2025-09-11T10:23:02.040Z
+ 2025-09-12T09:59:59.891Z
https://docs.axolotl.ai/docs/api/core.trainers.grpo.trainer.html
- 2025-09-11T10:23:01.858Z
+ 2025-09-12T09:59:59.714Z
https://docs.axolotl.ai/docs/api/common.architectures.html
- 2025-09-11T10:23:02.775Z
+ 2025-09-12T10:00:00.605Z
https://docs.axolotl.ai/docs/api/cli.utils.html
- 2025-09-11T10:23:01.766Z
+ 2025-09-12T09:59:59.622Z
https://docs.axolotl.ai/docs/api/prompt_strategies.orpo.chat_template.html
- 2025-09-11T10:23:02.136Z
+ 2025-09-12T09:59:59.988Z
https://docs.axolotl.ai/docs/api/core.training_args.html
- 2025-09-11T10:23:01.555Z
+ 2025-09-12T09:59:59.415Z
https://docs.axolotl.ai/docs/api/utils.schemas.datasets.html
- 2025-09-11T10:23:02.554Z
+ 2025-09-12T10:00:00.401Z
https://docs.axolotl.ai/docs/api/prompt_strategies.chat_template.html
- 2025-09-11T10:23:01.986Z
+ 2025-09-12T09:59:59.838Z
https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_w_system.html
- 2025-09-11T10:23:02.013Z
+ 2025-09-12T09:59:59.865Z
https://docs.axolotl.ai/docs/api/integrations.cut_cross_entropy.args.html
- 2025-09-11T10:23:02.755Z
+ 2025-09-12T10:00:00.586Z
https://docs.axolotl.ai/docs/api/common.const.html
- 2025-09-11T10:23:02.777Z
+ 2025-09-12T10:00:00.607Z
https://docs.axolotl.ai/docs/api/monkeypatch.lora_kernels.html
- 2025-09-11T10:23:02.299Z
+ 2025-09-12T10:00:00.153Z
https://docs.axolotl.ai/docs/api/kernels.quantize.html
- 2025-09-11T10:23:02.254Z
+ 2025-09-12T10:00:00.108Z
https://docs.axolotl.ai/docs/api/cli.config.html
- 2025-09-11T10:23:01.696Z
+ 2025-09-12T09:59:59.553Z
https://docs.axolotl.ai/docs/api/prompt_strategies.llama2_chat.html
- 2025-09-11T10:23:02.034Z
+ 2025-09-12T09:59:59.885Z
https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.zephyr.html
- 2025-09-11T10:23:02.095Z
+ 2025-09-12T09:59:59.947Z
https://docs.axolotl.ai/docs/api/kernels.geglu.html
- 2025-09-11T10:23:02.237Z
+ 2025-09-12T10:00:00.090Z
https://docs.axolotl.ai/docs/api/utils.freeze.html
- 2025-09-11T10:23:02.395Z
+ 2025-09-12T10:00:00.250Z
https://docs.axolotl.ai/docs/api/utils.callbacks.mlflow_.html
- 2025-09-11T10:23:02.876Z
+ 2025-09-12T10:00:00.707Z
https://docs.axolotl.ai/docs/api/core.trainers.grpo.sampler.html
- 2025-09-11T10:23:01.870Z
+ 2025-09-12T09:59:59.727Z
https://docs.axolotl.ai/docs/api/cli.utils.train.html
- 2025-09-11T10:23:01.806Z
+ 2025-09-12T09:59:59.662Z
https://docs.axolotl.ai/docs/api/cli.utils.load.html
- 2025-09-11T10:23:01.789Z
+ 2025-09-12T09:59:59.645Z
https://docs.axolotl.ai/docs/api/utils.collators.core.html
- 2025-09-11T10:23:02.794Z
+ 2025-09-12T10:00:00.624Z
https://docs.axolotl.ai/docs/api/integrations.liger.args.html
- 2025-09-11T10:23:02.767Z
+ 2025-09-12T10:00:00.597Z
https://docs.axolotl.ai/docs/api/cli.preprocess.html
- 2025-09-11T10:23:01.743Z
+ 2025-09-12T09:59:59.600Z
https://docs.axolotl.ai/docs/api/kernels.utils.html
- 2025-09-11T10:23:02.256Z
+ 2025-09-12T10:00:00.110Z
https://docs.axolotl.ai/docs/api/utils.callbacks.lisa.html
- 2025-09-11T10:23:02.872Z
+ 2025-09-12T10:00:00.703Z
https://docs.axolotl.ai/docs/api/monkeypatch.data.batch_dataset_fetcher.html
- 2025-09-11T10:23:02.335Z
+ 2025-09-12T10:00:00.189Z
https://docs.axolotl.ai/docs/api/core.trainers.mixins.optimizer.html
- 2025-09-11T10:23:01.919Z
+ 2025-09-12T09:59:59.770Z
https://docs.axolotl.ai/docs/api/loaders.processor.html
- 2025-09-11T10:23:01.891Z
+ 2025-09-12T09:59:59.748Z
https://docs.axolotl.ai/docs/api/core.trainers.utils.html
- 2025-09-11T10:23:01.872Z
+ 2025-09-12T09:59:59.728Z
https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.chat_template.html
- 2025-09-11T10:23:02.073Z
+ 2025-09-12T09:59:59.925Z
https://docs.axolotl.ai/docs/api/prompt_tokenizers.html
- 2025-09-11T10:23:01.517Z
+ 2025-09-12T09:59:59.378Z
https://docs.axolotl.ai/docs/api/utils.callbacks.comet_.html
- 2025-09-11T10:23:02.879Z
+ 2025-09-12T10:00:00.710Z
https://docs.axolotl.ai/docs/api/core.datasets.transforms.chat_builder.html
- 2025-09-11T10:23:01.595Z
+ 2025-09-12T09:59:59.455Z
https://docs.axolotl.ai/docs/api/utils.trainer.html
- 2025-09-11T10:23:02.413Z
+ 2025-09-12T10:00:00.267Z
https://docs.axolotl.ai/docs/api/cli.train.html
- 2025-09-11T10:23:01.640Z
+ 2025-09-12T09:59:59.497Z
https://docs.axolotl.ai/docs/api/integrations.lm_eval.args.html
- 2025-09-11T10:23:02.770Z
+ 2025-09-12T10:00:00.601Z
https://docs.axolotl.ai/docs/api/prompt_strategies.messages.chat.html
- 2025-09-11T10:23:02.072Z
+ 2025-09-12T09:59:59.923Z
https://docs.axolotl.ai/docs/api/utils.schemas.peft.html
- 2025-09-11T10:23:02.562Z
+ 2025-09-12T10:00:00.410Z
https://docs.axolotl.ai/docs/api/prompt_strategies.bradley_terry.llama3.html
- 2025-09-11T10:23:02.140Z
+ 2025-09-12T09:59:59.992Z
https://docs.axolotl.ai/docs/api/cli.utils.args.html
- 2025-09-11T10:23:01.778Z
+ 2025-09-12T09:59:59.634Z
https://docs.axolotl.ai/docs/api/prompt_strategies.stepwise_supervised.html
- 2025-09-11T10:23:02.050Z
+ 2025-09-12T09:59:59.902Z
https://docs.axolotl.ai/docs/api/common.datasets.html
- 2025-09-11T10:23:02.791Z
+ 2025-09-12T10:00:00.622Z
https://docs.axolotl.ai/docs/api/prompt_strategies.kto.user_defined.html
- 2025-09-11T10:23:02.116Z
+ 2025-09-12T09:59:59.967Z
https://docs.axolotl.ai/docs/api/train.html
- 2025-09-11T10:23:01.446Z
+ 2025-09-12T09:59:59.306Z
https://docs.axolotl.ai/docs/multi-node.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/input_output.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.651Z
https://docs.axolotl.ai/docs/ray-integration.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/getting-started.html
- 2025-09-11T10:19:40.169Z
+ 2025-09-12T09:55:59.649Z
https://docs.axolotl.ai/docs/optimizers.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/multi-gpu.html
- 2025-09-11T10:19:40.171Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/dataset_preprocessing.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/torchao.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/config-reference.html
- 2025-09-11T10:23:17.799Z
+ 2025-09-12T10:00:16.899Z
https://docs.axolotl.ai/docs/rlhf.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/docs/cli.html
- 2025-09-11T10:19:40.168Z
+ 2025-09-12T09:55:59.648Z
https://docs.axolotl.ai/docs/fsdp_qlora.html
- 2025-09-11T10:19:40.169Z
+ 2025-09-12T09:55:59.649Z
https://docs.axolotl.ai/docs/qat.html
- 2025-09-11T10:19:40.172Z
+ 2025-09-12T09:55:59.652Z
https://docs.axolotl.ai/src/axolotl/integrations/cut_cross_entropy/ACKNOWLEDGEMENTS.html
- 2025-09-11T10:19:40.193Z
+ 2025-09-12T09:55:59.674Z
https://docs.axolotl.ai/index.html
- 2025-09-11T10:19:40.188Z
+ 2025-09-12T09:55:59.669Z