diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml index a43dbac41..8692496f1 100644 --- a/.github/workflows/main.yml +++ b/.github/workflows/main.yml @@ -15,16 +15,11 @@ jobs: fail-fast: false matrix: include: - - cuda: 124 - cuda_version: 12.4.1 - python_version: "3.11" - pytorch: 2.5.1 - axolotl_extras: - cuda: 126 cuda_version: 12.6.3 python_version: "3.11" pytorch: 2.6.0 - axolotl_extras: vllm + axolotl_extras: - cuda: 126 cuda_version: 12.6.3 python_version: "3.11" @@ -87,11 +82,6 @@ jobs: strategy: matrix: include: - - cuda: 124 - cuda_version: 12.4.1 - python_version: "3.11" - pytorch: 2.5.1 - axolotl_extras: - cuda: 126 cuda_version: 12.6.3 python_version: "3.11" diff --git a/.nojekyll b/.nojekyll index 9869ad18c..fa762a488 100644 --- a/.nojekyll +++ b/.nojekyll @@ -1 +1 @@ -484f2284 \ No newline at end of file +610a007e \ No newline at end of file diff --git a/docs/config-reference.html b/docs/config-reference.html index 5365fec26..44e3203d1 100644 --- a/docs/config-reference.html +++ b/docs/config-reference.html @@ -1318,7 +1318,7 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true}); 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>=2.5.1 +# torch compile when torch>=2.6.0 torch_compile: Literal['auto'] | bool | None # Backend to use for torch.compile torch_compile_backend: str | None diff --git a/docs/custom_integrations.html b/docs/custom_integrations.html index 51843143c..e79bf9623 100644 --- a/docs/custom_integrations.html +++ b/docs/custom_integrations.html @@ -652,7 +652,7 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true}); kd_alpha: 0.9 kd_temperature: 1.0 -torch_compile: True # torch>=2.5.1, recommended to reduce vram +torch_compile: True # torch>=2.6.0, recommended to reduce vram datasets: - path: ... diff --git a/docs/docker.html b/docs/docker.html index 261411629..b4fcf619e 100644 --- a/docs/docker.html +++ b/docs/docker.html @@ -545,7 +545,6 @@ Important
  • main-base-py3.11-cu126-2.7.1
  • main-base-py3.11-cu126-2.6.0
  • main-base-py3.11-cu124-2.6.0
    • -
  • main-base-py3.11-cu124-2.5.1
    • @@ -589,10 +588,9 @@ Tip
  • main-py3.11-cu126-2.7.1
  • main-py3.11-cu126-2.6.0
  • main-py3.11-cu124-2.6.0
    • -
  • main-py3.11-cu124-2.5.1
  • main-latest
  • main-20250303-py3.11-cu124-2.6.0
    • -
  • main-20250303-py3.11-cu124-2.5.1
    • +
  • main-20250303-py3.11-cu126-2.6.0
  • 0.10.1
    • diff --git a/docs/installation.html b/docs/installation.html index 65afa2c76..aec81d44f 100644 --- a/docs/installation.html +++ b/docs/installation.html @@ -521,7 +521,7 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
    diff --git a/index.html b/index.html index ce99951f6..c5532bff8 100644 --- a/index.html +++ b/index.html @@ -542,7 +542,7 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});

    Installation

    diff --git a/search.json b/search.json index 35aecbb42..bc13ab0f4 100644 --- a/search.json +++ b/search.json @@ -169,7 +169,7 @@ "href": "docs/docker.html#base", "title": "Docker", "section": "Base", - "text": "Base\nThe base image is the most minimal image that can install Axolotl. It is based on the nvidia/cuda image. It includes python, torch, git, git-lfs, awscli, pydantic, and more.\n\nImage\naxolotlai/axolotl-base\nLink: Docker Hub\n\n\nTags format\nmain-base-py{python_version}-cu{cuda_version}-{pytorch_version}\nTags examples:\n\nmain-base-py3.11-cu128-2.7.1\nmain-base-py3.11-cu126-2.7.1\nmain-base-py3.11-cu126-2.6.0\nmain-base-py3.11-cu124-2.6.0\nmain-base-py3.11-cu124-2.5.1", + "text": "Base\nThe base image is the most minimal image that can install Axolotl. It is based on the nvidia/cuda image. It includes python, torch, git, git-lfs, awscli, pydantic, and more.\n\nImage\naxolotlai/axolotl-base\nLink: Docker Hub\n\n\nTags format\nmain-base-py{python_version}-cu{cuda_version}-{pytorch_version}\nTags examples:\n\nmain-base-py3.11-cu128-2.7.1\nmain-base-py3.11-cu126-2.7.1\nmain-base-py3.11-cu126-2.6.0\nmain-base-py3.11-cu124-2.6.0", "crumbs": [ "Deployments", "Docker" @@ -180,7 +180,7 @@ "href": "docs/docker.html#main", "title": "Docker", "section": "Main", - "text": "Main\nThe main image is the image that is used to run Axolotl. It is based on the axolotlai/axolotl-base image and includes the Axolotl codebase, dependencies, and more.\n\nImage\naxolotlai/axolotl\nLink: Docker Hub\n\n\nTags format\n# on push to main\nmain-py{python_version}-cu{cuda_version}-{pytorch_version}\n\n# latest main (currently torch 2.6.0, python 3.11, cuda 12.4)\nmain-latest\n\n# nightly build\n{branch}-{date_in_YYYYMMDD}-py{python_version}-cu{cuda_version}-{pytorch_version}\n\n# tagged release\n{version}\n\n\n\n\n\n\nTip\n\n\n\nThere may be some extra tags appended to the image, like -vllm which installs those packages.\n\n\nTags examples:\n\nmain-py3.11-cu128-2.7.1\nmain-py3.11-cu126-2.7.1\nmain-py3.11-cu126-2.6.0\nmain-py3.11-cu124-2.6.0\nmain-py3.11-cu124-2.5.1\nmain-latest\nmain-20250303-py3.11-cu124-2.6.0\nmain-20250303-py3.11-cu124-2.5.1\n0.10.1", + "text": "Main\nThe main image is the image that is used to run Axolotl. It is based on the axolotlai/axolotl-base image and includes the Axolotl codebase, dependencies, and more.\n\nImage\naxolotlai/axolotl\nLink: Docker Hub\n\n\nTags format\n# on push to main\nmain-py{python_version}-cu{cuda_version}-{pytorch_version}\n\n# latest main (currently torch 2.6.0, python 3.11, cuda 12.4)\nmain-latest\n\n# nightly build\n{branch}-{date_in_YYYYMMDD}-py{python_version}-cu{cuda_version}-{pytorch_version}\n\n# tagged release\n{version}\n\n\n\n\n\n\nTip\n\n\n\nThere may be some extra tags appended to the image, like -vllm which installs those packages.\n\n\nTags examples:\n\nmain-py3.11-cu128-2.7.1\nmain-py3.11-cu126-2.7.1\nmain-py3.11-cu126-2.6.0\nmain-py3.11-cu124-2.6.0\nmain-latest\nmain-20250303-py3.11-cu124-2.6.0\nmain-20250303-py3.11-cu126-2.6.0\n0.10.1", "crumbs": [ "Deployments", "Docker" @@ -367,7 +367,7 @@ "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\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 # 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 # 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 # 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\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 # For `completion` datsets only, uses the provided field instead of `text` column\n field: 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 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\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 # For `completion` datsets only, uses the provided field instead of `text` column\n field: 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 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\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# 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\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 # For `completion` datsets only, uses the provided field instead of `text` column\n field: 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 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\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.\ndataset_processes: int | None = 4\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\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\nfp8: 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\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# 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\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# whether to concatenate samples during pretraining\npretraining_sample_concatenation: 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 = 10000\n# whether to prevent cross attention for packed sequences during pretraining\npretrain_multipack_attn: bool | None = True\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 QKV into a single operation\nflash_attn_fuse_qkv: 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\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\nllama4_linearized_experts: bool | None\n\n# Deepspeed config path. e.g., deepspeed_configs/zero3.json\ndeepspeed: str | dict[str, Any] | None\n# FSDP configuration\nfsdp: list[str] | None\n# FSDP configuration options\nfsdp_config: dict[str, Any] | 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# 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.\nsequence_parallel_degree: 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\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>=2.5.1\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# 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# 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# bool of whether to include tokens trainer per second in the training metrics. This\n# iterates over the entire dataset once, so it takes some time.\ninclude_tokens_per_second: bool | None\n\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\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# Trust remote code for untrusted source\ntrust_remote_code: bool | 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\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\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# Number of steps per ReLoRA restart\nrelora_steps: int | None\n# Number of per-restart warmup steps\nrelora_warmup_steps: int | None\n# Number of anneal steps for each relora cycle\nrelora_anneal_steps: int | 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# 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# 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\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 # 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 # 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 # 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\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 # For `completion` datsets only, uses the provided field instead of `text` column\n field: 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 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\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 # For `completion` datsets only, uses the provided field instead of `text` column\n field: 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 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\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# 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\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 # For `completion` datsets only, uses the provided field instead of `text` column\n field: 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 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\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.\ndataset_processes: int | None = 4\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\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\nfp8: 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\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# 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\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# whether to concatenate samples during pretraining\npretraining_sample_concatenation: 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 = 10000\n# whether to prevent cross attention for packed sequences during pretraining\npretrain_multipack_attn: bool | None = True\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 QKV into a single operation\nflash_attn_fuse_qkv: 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\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\nllama4_linearized_experts: bool | None\n\n# Deepspeed config path. e.g., deepspeed_configs/zero3.json\ndeepspeed: str | dict[str, Any] | None\n# FSDP configuration\nfsdp: list[str] | None\n# FSDP configuration options\nfsdp_config: dict[str, Any] | 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# 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.\nsequence_parallel_degree: 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\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# 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# 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# bool of whether to include tokens trainer per second in the training metrics. This\n# iterates over the entire dataset once, so it takes some time.\ninclude_tokens_per_second: bool | None\n\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\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# Trust remote code for untrusted source\ntrust_remote_code: bool | 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\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\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# Number of steps per ReLoRA restart\nrelora_steps: int | None\n# Number of per-restart warmup steps\nrelora_warmup_steps: int | None\n# Number of anneal steps for each relora cycle\nrelora_anneal_steps: int | 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# 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# 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" @@ -1756,7 +1756,7 @@ "href": "index.html#quick-start", "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 ≥2.5.1\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\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.", + "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\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\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" ] @@ -3097,7 +3097,7 @@ "href": "docs/custom_integrations.html#knowledge-distillation-kd", "title": "Custom Integrations", "section": "Knowledge Distillation (KD)", - "text": "Knowledge Distillation (KD)\n\nUsage\nplugins:\n - \"axolotl.integrations.kd.KDPlugin\"\n\nkd_trainer: True\nkd_ce_alpha: 0.1\nkd_alpha: 0.9\nkd_temperature: 1.0\n\ntorch_compile: True # torch>=2.5.1, recommended to reduce vram\n\ndatasets:\n - path: ...\n type: \"axolotl.integrations.kd.chat_template\"\n field_messages: \"messages_combined\"\n logprobs_field: \"llm_text_generation_vllm_logprobs\" # for kd only, field of logprobs\nAn example dataset can be found at axolotl-ai-co/evolkit-logprobs-pipeline-75k-v2-sample\nPlease see reference here", + "text": "Knowledge Distillation (KD)\n\nUsage\nplugins:\n - \"axolotl.integrations.kd.KDPlugin\"\n\nkd_trainer: True\nkd_ce_alpha: 0.1\nkd_alpha: 0.9\nkd_temperature: 1.0\n\ntorch_compile: True # torch>=2.6.0, recommended to reduce vram\n\ndatasets:\n - path: ...\n type: \"axolotl.integrations.kd.chat_template\"\n field_messages: \"messages_combined\"\n logprobs_field: \"llm_text_generation_vllm_logprobs\" # for kd only, field of logprobs\nAn example dataset can be found at axolotl-ai-co/evolkit-logprobs-pipeline-75k-v2-sample\nPlease see reference here", "crumbs": [ "Advanced Features", "Custom Integrations" @@ -3361,7 +3361,7 @@ "href": "docs/installation.html#sec-requirements", "title": "Installation", "section": "1 Requirements", - "text": "1 Requirements\n\nNVIDIA GPU (Ampere architecture or newer for bf16 and Flash Attention) or AMD GPU\nPython ≥3.11\nPyTorch ≥2.5.1", + "text": "1 Requirements\n\nNVIDIA GPU (Ampere architecture or newer for bf16 and Flash Attention) or AMD GPU\nPython ≥3.11\nPyTorch ≥2.6.0", "crumbs": [ "Getting Started", "Installation" diff --git a/sitemap.xml b/sitemap.xml index 911edaf1a..bfc9723d5 100644 --- a/sitemap.xml +++ b/sitemap.xml @@ -2,758 +2,758 @@ https://docs.axolotl.ai/docs/unsloth.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/dataset-formats/conversation.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/dataset-formats/stepwise_supervised.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/dataset-formats/tokenized.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/mac.html - 2025-07-09T12:43:49.326Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/nccl.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/multi-node.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/docker.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/lr_groups.html - 2025-07-09T12:43:49.326Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/inference.html - 2025-07-09T12:43:49.326Z + 2025-07-09T13:22:46.805Z https://docs.axolotl.ai/docs/cli.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/config-reference.html - 2025-07-09T12:47:25.276Z + 2025-07-09T13:26:17.755Z https://docs.axolotl.ai/docs/multi-gpu.html - 2025-07-09T12:43:49.326Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/debugging.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/multimodal.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/api/cli.sweeps.html - 2025-07-09T12:47:11.129Z + 2025-07-09T13:26:03.594Z https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.llama3.html - 2025-07-09T12:47:11.457Z + 2025-07-09T13:26:03.923Z https://docs.axolotl.ai/docs/api/utils.schedulers.html - 2025-07-09T12:47:11.843Z + 2025-07-09T13:26:04.310Z https://docs.axolotl.ai/docs/api/monkeypatch.llama_attn_hijack_xformers.html - 2025-07-09T12:47:11.656Z + 2025-07-09T13:26:04.123Z https://docs.axolotl.ai/docs/api/cli.cloud.modal_.html - 2025-07-09T12:47:11.178Z + 2025-07-09T13:26:03.643Z https://docs.axolotl.ai/docs/api/kernels.geglu.html - 2025-07-09T12:47:11.608Z + 2025-07-09T13:26:04.076Z https://docs.axolotl.ai/docs/api/core.trainers.utils.html - 2025-07-09T12:47:11.251Z + 2025-07-09T13:26:03.716Z https://docs.axolotl.ai/docs/api/core.datasets.chat.html - 2025-07-09T12:47:10.977Z + 2025-07-09T13:26:03.432Z https://docs.axolotl.ai/docs/api/utils.schemas.peft.html - 2025-07-09T12:47:11.957Z + 2025-07-09T13:26:04.424Z https://docs.axolotl.ai/docs/api/monkeypatch.btlm_attn_hijack_flash.html - 2025-07-09T12:47:11.719Z + 2025-07-09T13:26:04.187Z https://docs.axolotl.ai/docs/api/prompt_strategies.chat_template.html - 2025-07-09T12:47:11.359Z + 2025-07-09T13:26:03.825Z https://docs.axolotl.ai/docs/api/prompt_strategies.kto.user_defined.html - 2025-07-09T12:47:11.490Z + 2025-07-09T13:26:03.956Z https://docs.axolotl.ai/docs/api/cli.cloud.base.html - 2025-07-09T12:47:11.172Z + 2025-07-09T13:26:03.636Z https://docs.axolotl.ai/docs/api/kernels.swiglu.html - 2025-07-09T12:47:11.619Z + 2025-07-09T13:26:04.086Z https://docs.axolotl.ai/docs/api/prompt_strategies.stepwise_supervised.html - 2025-07-09T12:47:11.424Z + 2025-07-09T13:26:03.890Z https://docs.axolotl.ai/docs/api/prompt_strategies.bradley_terry.llama3.html - 2025-07-09T12:47:11.514Z + 2025-07-09T13:26:03.980Z https://docs.axolotl.ai/docs/api/prompt_strategies.completion.html - 2025-07-09T12:47:11.413Z + 2025-07-09T13:26:03.879Z https://docs.axolotl.ai/docs/api/kernels.utils.html - 2025-07-09T12:47:11.628Z + 2025-07-09T13:26:04.095Z https://docs.axolotl.ai/docs/api/common.datasets.html - 2025-07-09T12:47:12.170Z + 2025-07-09T13:26:04.635Z https://docs.axolotl.ai/docs/api/utils.schemas.datasets.html - 2025-07-09T12:47:11.948Z + 2025-07-09T13:26:04.415Z https://docs.axolotl.ai/docs/api/core.builders.rl.html - 2025-07-09T12:47:10.932Z + 2025-07-09T13:26:03.386Z https://docs.axolotl.ai/docs/api/evaluate.html - 2025-07-09T12:47:10.835Z + 2025-07-09T13:26:03.291Z https://docs.axolotl.ai/docs/api/kernels.quantize.html - 2025-07-09T12:47:11.627Z + 2025-07-09T13:26:04.094Z https://docs.axolotl.ai/docs/api/monkeypatch.llama_attn_hijack_flash.html - 2025-07-09T12:47:11.655Z + 2025-07-09T13:26:04.122Z https://docs.axolotl.ai/docs/api/core.trainers.mixins.rng_state_loader.html - 2025-07-09T12:47:11.295Z + 2025-07-09T13:26:03.760Z https://docs.axolotl.ai/docs/api/integrations.base.html - 2025-07-09T12:47:12.130Z + 2025-07-09T13:26:04.595Z https://docs.axolotl.ai/docs/api/cli.merge_lora.html - 2025-07-09T12:47:11.102Z + 2025-07-09T13:26:03.557Z https://docs.axolotl.ai/docs/api/cli.merge_sharded_fsdp_weights.html - 2025-07-09T12:47:11.114Z + 2025-07-09T13:26:03.569Z https://docs.axolotl.ai/docs/api/monkeypatch.transformers_fa_utils.html - 2025-07-09T12:47:11.736Z + 2025-07-09T13:26:04.204Z https://docs.axolotl.ai/docs/api/prompt_strategies.llama2_chat.html - 2025-07-09T12:47:11.407Z + 2025-07-09T13:26:03.873Z https://docs.axolotl.ai/docs/api/utils.collators.mm_chat.html - 2025-07-09T12:47:12.200Z + 2025-07-09T13:26:04.664Z https://docs.axolotl.ai/docs/api/utils.data.sft.html - 2025-07-09T12:47:11.883Z + 2025-07-09T13:26:04.349Z https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_instruct.html - 2025-07-09T12:47:11.374Z + 2025-07-09T13:26:03.840Z https://docs.axolotl.ai/docs/api/integrations.liger.args.html - 2025-07-09T12:47:12.146Z + 2025-07-09T13:26:04.610Z https://docs.axolotl.ai/docs/api/monkeypatch.mistral_attn_hijack_flash.html - 2025-07-09T12:47:11.671Z + 2025-07-09T13:26:04.138Z https://docs.axolotl.ai/docs/api/cli.vllm_serve.html - 2025-07-09T12:47:11.169Z + 2025-07-09T13:26:03.633Z https://docs.axolotl.ai/docs/api/monkeypatch.utils.html - 2025-07-09T12:47:11.718Z + 2025-07-09T13:26:04.185Z https://docs.axolotl.ai/docs/api/loaders.patch_manager.html - 2025-07-09T12:47:11.284Z + 2025-07-09T13:26:03.750Z https://docs.axolotl.ai/docs/api/utils.schemas.integrations.html - 2025-07-09T12:47:11.978Z + 2025-07-09T13:26:04.444Z https://docs.axolotl.ai/docs/api/utils.callbacks.perplexity.html - 2025-07-09T12:47:12.248Z + 2025-07-09T13:26:04.712Z https://docs.axolotl.ai/docs/api/cli.utils.html - 2025-07-09T12:47:11.161Z + 2025-07-09T13:26:03.626Z https://docs.axolotl.ai/docs/api/utils.schemas.config.html - 2025-07-09T12:47:11.918Z + 2025-07-09T13:26:04.385Z https://docs.axolotl.ai/docs/api/prompt_strategies.input_output.html - 2025-07-09T12:47:11.419Z + 2025-07-09T13:26:03.885Z https://docs.axolotl.ai/docs/api/utils.distributed.html - 2025-07-09T12:47:11.863Z + 2025-07-09T13:26:04.330Z https://docs.axolotl.ai/docs/api/monkeypatch.gradient_checkpointing.offload_disk.html - 2025-07-09T12:47:11.770Z + 2025-07-09T13:26:04.238Z https://docs.axolotl.ai/docs/api/monkeypatch.trainer_fsdp_optim.html - 2025-07-09T12:47:11.730Z + 2025-07-09T13:26:04.198Z https://docs.axolotl.ai/docs/api/core.builders.base.html - 2025-07-09T12:47:10.919Z + 2025-07-09T13:26:03.374Z https://docs.axolotl.ai/docs/api/core.trainers.trl.html - 2025-07-09T12:47:11.210Z + 2025-07-09T13:26:03.675Z https://docs.axolotl.ai/docs/api/cli.evaluate.html - 2025-07-09T12:47:11.033Z + 2025-07-09T13:26:03.489Z https://docs.axolotl.ai/docs/api/utils.optimizers.adopt.html - 2025-07-09T12:47:11.874Z + 2025-07-09T13:26:04.341Z https://docs.axolotl.ai/docs/api/utils.callbacks.qat.html - 2025-07-09T12:47:12.267Z + 2025-07-09T13:26:04.731Z https://docs.axolotl.ai/docs/api/core.trainers.dpo.trainer.html - 2025-07-09T12:47:11.226Z + 2025-07-09T13:26:03.692Z https://docs.axolotl.ai/docs/api/core.chat.format.shared.html - 2025-07-09T12:47:10.972Z + 2025-07-09T13:26:03.427Z https://docs.axolotl.ai/docs/api/monkeypatch.relora.html - 2025-07-09T12:47:11.679Z + 2025-07-09T13:26:04.147Z https://docs.axolotl.ai/docs/api/cli.config.html - 2025-07-09T12:47:11.078Z + 2025-07-09T13:26:03.533Z https://docs.axolotl.ai/docs/api/cli.preprocess.html - 2025-07-09T12:47:11.123Z + 2025-07-09T13:26:03.585Z https://docs.axolotl.ai/docs/api/core.trainers.base.html - 2025-07-09T12:47:11.194Z + 2025-07-09T13:26:03.658Z https://docs.axolotl.ai/docs/api/convert.html - 2025-07-09T12:47:10.860Z + 2025-07-09T13:26:03.315Z https://docs.axolotl.ai/docs/api/prompt_strategies.pygmalion.html - 2025-07-09T12:47:11.441Z + 2025-07-09T13:26:03.907Z https://docs.axolotl.ai/docs/api/utils.schemas.trl.html - 2025-07-09T12:47:11.960Z + 2025-07-09T13:26:04.427Z https://docs.axolotl.ai/docs/api/cli.args.html - 2025-07-09T12:47:11.053Z + 2025-07-09T13:26:03.509Z https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.chat_template.html - 2025-07-09T12:47:11.447Z + 2025-07-09T13:26:03.913Z https://docs.axolotl.ai/docs/api/loaders.constants.html - 2025-07-09T12:47:11.286Z + 2025-07-09T13:26:03.751Z https://docs.axolotl.ai/docs/api/logging_config.html - 2025-07-09T12:47:10.913Z + 2025-07-09T13:26:03.367Z https://docs.axolotl.ai/docs/api/cli.inference.html - 2025-07-09T12:47:11.093Z + 2025-07-09T13:26:03.548Z https://docs.axolotl.ai/docs/api/utils.ctx_managers.sequence_parallel.html - 2025-07-09T12:47:11.325Z + 2025-07-09T13:26:03.790Z https://docs.axolotl.ai/docs/api/integrations.spectrum.args.html - 2025-07-09T12:47:12.152Z + 2025-07-09T13:26:04.617Z https://docs.axolotl.ai/docs/api/utils.schemas.training.html - 2025-07-09T12:47:11.930Z + 2025-07-09T13:26:04.397Z https://docs.axolotl.ai/docs/api/prompt_strategies.orcamini.html - 2025-07-09T12:47:11.434Z + 2025-07-09T13:26:03.900Z https://docs.axolotl.ai/docs/api/utils.freeze.html - 2025-07-09T12:47:11.800Z + 2025-07-09T13:26:04.268Z https://docs.axolotl.ai/docs/api/loaders.tokenizer.html - 2025-07-09T12:47:11.269Z + 2025-07-09T13:26:03.735Z https://docs.axolotl.ai/docs/api/utils.bench.html - 2025-07-09T12:47:11.792Z + 2025-07-09T13:26:04.260Z https://docs.axolotl.ai/docs/api/utils.quantization.html - 2025-07-09T12:47:11.904Z + 2025-07-09T13:26:04.371Z https://docs.axolotl.ai/docs/batch_vs_grad.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/input_output.html - 2025-07-09T12:43:49.326Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/sequence_parallelism.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/reward_modelling.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/index.html - 2025-07-09T12:43:49.342Z + 2025-07-09T13:22:46.821Z https://docs.axolotl.ai/src/axolotl/integrations/LICENSE.html - 2025-07-09T12:43:49.345Z + 2025-07-09T13:22:46.825Z https://docs.axolotl.ai/FAQS.html - 2025-07-09T12:43:49.321Z + 2025-07-09T13:22:46.801Z https://docs.axolotl.ai/src/axolotl/integrations/cut_cross_entropy/ACKNOWLEDGEMENTS.html - 2025-07-09T12:43:49.346Z + 2025-07-09T13:22:46.825Z https://docs.axolotl.ai/TODO.html - 2025-07-09T12:43:49.321Z + 2025-07-09T13:22:46.801Z https://docs.axolotl.ai/examples/colab-notebooks/colab-axolotl-example.html - 2025-07-09T12:43:49.329Z + 2025-07-09T13:22:46.808Z https://docs.axolotl.ai/docs/torchao.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/ray-integration.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/quantize.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/qat.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/api/utils.lora.html - 2025-07-09T12:47:11.783Z + 2025-07-09T13:26:04.251Z https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_w_system.html - 2025-07-09T12:47:11.386Z + 2025-07-09T13:26:03.852Z https://docs.axolotl.ai/docs/api/monkeypatch.stablelm_attn_hijack_flash.html - 2025-07-09T12:47:11.726Z + 2025-07-09T13:26:04.194Z https://docs.axolotl.ai/docs/api/utils.collators.core.html - 2025-07-09T12:47:12.173Z + 2025-07-09T13:26:04.637Z https://docs.axolotl.ai/docs/api/prompt_strategies.metharme.html - 2025-07-09T12:47:11.431Z + 2025-07-09T13:26:03.896Z https://docs.axolotl.ai/docs/api/utils.callbacks.profiler.html - 2025-07-09T12:47:12.251Z + 2025-07-09T13:26:04.716Z https://docs.axolotl.ai/docs/api/utils.data.pretraining.html - 2025-07-09T12:47:11.876Z + 2025-07-09T13:26:04.343Z https://docs.axolotl.ai/docs/api/utils.callbacks.lisa.html - 2025-07-09T12:47:12.253Z + 2025-07-09T13:26:04.717Z https://docs.axolotl.ai/docs/api/utils.trainer.html - 2025-07-09T12:47:11.818Z + 2025-07-09T13:26:04.286Z https://docs.axolotl.ai/docs/api/integrations.cut_cross_entropy.args.html - 2025-07-09T12:47:12.134Z + 2025-07-09T13:26:04.598Z https://docs.axolotl.ai/docs/api/utils.schemas.model.html - 2025-07-09T12:47:11.925Z + 2025-07-09T13:26:04.392Z https://docs.axolotl.ai/docs/api/monkeypatch.data.batch_dataset_fetcher.html - 2025-07-09T12:47:11.739Z + 2025-07-09T13:26:04.207Z https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.zephyr.html - 2025-07-09T12:47:11.469Z + 2025-07-09T13:26:03.935Z https://docs.axolotl.ai/docs/api/datasets.html - 2025-07-09T12:47:10.846Z + 2025-07-09T13:26:03.301Z https://docs.axolotl.ai/docs/api/utils.schemas.enums.html - 2025-07-09T12:47:11.989Z + 2025-07-09T13:26:04.455Z https://docs.axolotl.ai/docs/api/integrations.kd.trainer.html - 2025-07-09T12:47:12.142Z + 2025-07-09T13:26:04.607Z https://docs.axolotl.ai/docs/api/monkeypatch.lora_kernels.html - 2025-07-09T12:47:11.709Z + 2025-07-09T13:26:04.177Z https://docs.axolotl.ai/docs/api/utils.collators.batching.html - 2025-07-09T12:47:12.192Z + 2025-07-09T13:26:04.656Z https://docs.axolotl.ai/docs/api/core.trainers.grpo.sampler.html - 2025-07-09T12:47:11.250Z + 2025-07-09T13:26:03.715Z https://docs.axolotl.ai/docs/api/prompt_strategies.base.html - 2025-07-09T12:47:11.327Z + 2025-07-09T13:26:03.792Z https://docs.axolotl.ai/docs/api/monkeypatch.multipack.html - 2025-07-09T12:47:11.673Z + 2025-07-09T13:26:04.140Z https://docs.axolotl.ai/docs/api/prompt_strategies.orpo.chat_template.html - 2025-07-09T12:47:11.510Z + 2025-07-09T13:26:03.976Z https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.passthrough.html - 2025-07-09T12:47:11.472Z + 2025-07-09T13:26:03.938Z https://docs.axolotl.ai/docs/api/core.chat.format.chatml.html - 2025-07-09T12:47:10.969Z + 2025-07-09T13:26:03.424Z https://docs.axolotl.ai/docs/api/core.trainers.mixins.scheduler.html - 2025-07-09T12:47:11.301Z + 2025-07-09T13:26:03.767Z https://docs.axolotl.ai/docs/api/utils.model_shard_quant.html - 2025-07-09T12:47:11.789Z + 2025-07-09T13:26:04.257Z https://docs.axolotl.ai/docs/api/prompt_strategies.kto.chatml.html - 2025-07-09T12:47:11.488Z + 2025-07-09T13:26:03.954Z https://docs.axolotl.ai/docs/api/utils.tokenization.html - 2025-07-09T12:47:11.777Z + 2025-07-09T13:26:04.245Z https://docs.axolotl.ai/docs/api/loaders.model.html - 2025-07-09T12:47:11.261Z + 2025-07-09T13:26:03.726Z https://docs.axolotl.ai/docs/api/utils.callbacks.mlflow_.html - 2025-07-09T12:47:12.256Z + 2025-07-09T13:26:04.721Z https://docs.axolotl.ai/docs/api/core.trainers.grpo.trainer.html - 2025-07-09T12:47:11.237Z + 2025-07-09T13:26:03.702Z https://docs.axolotl.ai/docs/api/cli.main.html - 2025-07-09T12:47:11.016Z + 2025-07-09T13:26:03.472Z https://docs.axolotl.ai/docs/api/utils.callbacks.comet_.html - 2025-07-09T12:47:12.260Z + 2025-07-09T13:26:04.724Z https://docs.axolotl.ai/docs/api/utils.chat_templates.html - 2025-07-09T12:47:11.778Z + 2025-07-09T13:26:04.246Z https://docs.axolotl.ai/docs/api/utils.schemas.utils.html - 2025-07-09T12:47:11.994Z + 2025-07-09T13:26:04.460Z https://docs.axolotl.ai/docs/api/common.architectures.html - 2025-07-09T12:47:12.154Z + 2025-07-09T13:26:04.618Z https://docs.axolotl.ai/docs/api/monkeypatch.llama_expand_mask.html - 2025-07-09T12:47:11.681Z + 2025-07-09T13:26:04.148Z https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_chat.html - 2025-07-09T12:47:11.373Z + 2025-07-09T13:26:03.839Z https://docs.axolotl.ai/docs/api/utils.samplers.multipack.html - 2025-07-09T12:47:12.241Z + 2025-07-09T13:26:04.706Z https://docs.axolotl.ai/docs/api/integrations.grokfast.optimizer.html - 2025-07-09T12:47:12.135Z + 2025-07-09T13:26:04.599Z https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.chatml.html - 2025-07-09T12:47:11.468Z + 2025-07-09T13:26:03.933Z https://docs.axolotl.ai/docs/api/monkeypatch.mixtral.html - 2025-07-09T12:47:11.741Z + 2025-07-09T13:26:04.209Z https://docs.axolotl.ai/docs/api/train.html - 2025-07-09T12:47:10.825Z + 2025-07-09T13:26:03.280Z https://docs.axolotl.ai/docs/api/monkeypatch.llama_patch_multipack.html - 2025-07-09T12:47:11.721Z + 2025-07-09T13:26:04.188Z https://docs.axolotl.ai/docs/api/index.html - 2025-07-09T12:47:10.762Z + 2025-07-09T13:26:03.218Z https://docs.axolotl.ai/docs/api/loaders.adapter.html - 2025-07-09T12:47:11.276Z + 2025-07-09T13:26:03.742Z https://docs.axolotl.ai/docs/api/utils.schemas.multimodal.html - 2025-07-09T12:47:11.966Z + 2025-07-09T13:26:04.432Z https://docs.axolotl.ai/docs/api/kernels.lora.html - 2025-07-09T12:47:11.598Z + 2025-07-09T13:26:04.065Z https://docs.axolotl.ai/docs/api/prompt_strategies.kto.llama3.html - 2025-07-09T12:47:11.480Z + 2025-07-09T13:26:03.946Z https://docs.axolotl.ai/docs/api/cli.checks.html - 2025-07-09T12:47:11.060Z + 2025-07-09T13:26:03.515Z https://docs.axolotl.ai/docs/api/cli.quantize.html - 2025-07-09T12:47:11.183Z + 2025-07-09T13:26:03.648Z https://docs.axolotl.ai/docs/api/integrations.lm_eval.args.html - 2025-07-09T12:47:12.149Z + 2025-07-09T13:26:04.613Z https://docs.axolotl.ai/docs/api/core.chat.messages.html - 2025-07-09T12:47:10.968Z + 2025-07-09T13:26:03.422Z https://docs.axolotl.ai/docs/api/core.builders.causal.html - 2025-07-09T12:47:10.924Z + 2025-07-09T13:26:03.378Z https://docs.axolotl.ai/docs/api/core.trainers.relora.html - 2025-07-09T12:47:11.220Z + 2025-07-09T13:26:03.685Z https://docs.axolotl.ai/docs/api/models.mamba.modeling_mamba.html - 2025-07-09T12:47:12.171Z + 2025-07-09T13:26:04.636Z https://docs.axolotl.ai/docs/api/monkeypatch.gradient_checkpointing.offload_cpu.html - 2025-07-09T12:47:11.744Z + 2025-07-09T13:26:04.212Z https://docs.axolotl.ai/docs/api/core.trainers.mamba.html - 2025-07-09T12:47:11.215Z + 2025-07-09T13:26:03.680Z https://docs.axolotl.ai/docs/api/core.datasets.transforms.chat_builder.html - 2025-07-09T12:47:10.985Z + 2025-07-09T13:26:03.440Z https://docs.axolotl.ai/docs/api/loaders.processor.html - 2025-07-09T12:47:11.271Z + 2025-07-09T13:26:03.736Z https://docs.axolotl.ai/docs/api/core.chat.format.llama3x.html - 2025-07-09T12:47:10.971Z + 2025-07-09T13:26:03.425Z https://docs.axolotl.ai/docs/api/prompt_strategies.messages.chat.html - 2025-07-09T12:47:11.445Z + 2025-07-09T13:26:03.911Z https://docs.axolotl.ai/docs/api/cli.train.html - 2025-07-09T12:47:11.025Z + 2025-07-09T13:26:03.481Z https://docs.axolotl.ai/docs/api/core.trainers.mixins.optimizer.html - 2025-07-09T12:47:11.291Z + 2025-07-09T13:26:03.757Z https://docs.axolotl.ai/docs/api/utils.collators.mamba.html - 2025-07-09T12:47:12.195Z + 2025-07-09T13:26:04.660Z https://docs.axolotl.ai/docs/api/monkeypatch.unsloth_.html - 2025-07-09T12:47:11.738Z + 2025-07-09T13:26:04.206Z https://docs.axolotl.ai/docs/api/utils.dict.html - 2025-07-09T12:47:11.867Z + 2025-07-09T13:26:04.333Z https://docs.axolotl.ai/docs/api/prompt_strategies.user_defined.html - 2025-07-09T12:47:11.394Z + 2025-07-09T13:26:03.861Z https://docs.axolotl.ai/docs/api/core.training_args.html - 2025-07-09T12:47:10.945Z + 2025-07-09T13:26:03.399Z https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.user_defined.html - 2025-07-09T12:47:11.470Z + 2025-07-09T13:26:03.936Z https://docs.axolotl.ai/docs/api/prompt_tokenizers.html - 2025-07-09T12:47:10.903Z + 2025-07-09T13:26:03.358Z https://docs.axolotl.ai/docs/api/common.const.html - 2025-07-09T12:47:12.155Z + 2025-07-09T13:26:04.620Z https://docs.axolotl.ai/docs/fsdp_qlora.html - 2025-07-09T12:43:49.324Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/custom_integrations.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/getting-started.html - 2025-07-09T12:43:49.324Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/faq.html - 2025-07-09T12:43:49.324Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/lora_optims.html - 2025-07-09T12:43:49.326Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/rlhf.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/amd_hpc.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/installation.html - 2025-07-09T12:43:49.326Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/multipack.html - 2025-07-09T12:43:49.327Z + 2025-07-09T13:22:46.806Z https://docs.axolotl.ai/docs/dataset_preprocessing.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/dataset_loading.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.803Z https://docs.axolotl.ai/docs/dataset-formats/inst_tune.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/dataset-formats/template_free.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/dataset-formats/index.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z https://docs.axolotl.ai/docs/dataset-formats/pretraining.html - 2025-07-09T12:43:49.323Z + 2025-07-09T13:22:46.802Z