core.training_args.AxolotlCPOConfig(
- model_type=None,
- lr_quadratic_warmup=False,
- pretraining=False,
- sample_packing=False,
- sample_packing_sequentially=False,
- multipack_real_batches=False,
- eval_sample_packing=None,
- sample_packing_efficiency=1.0,
- sample_packing_bin_size=200,
- sample_packing_group_size=100000,
- max_seq_length=2048,
- dataset_num_proc=None,
- relora_steps=None,
- relora_warmup_steps=None,
- relora_anneal_steps=None,
- relora_prune_ratio=0.9,
- bench_split='eval',
- bench_dataset='pharaouk/dharma-1/dharma_1_mini.json',
- do_bench_eval=False,
- do_causal_lm_eval=False,
- max_bench_samples=None,
- bench_source_max_len=2048,
- dataloader_prefetch_factor=None,
- cosine_min_lr_ratio=None,
- cosine_constant_lr_ratio=None,
- loraplus_lr_ratio=None,
- loraplus_lr_embedding=1e-06,
- embedding_lr_scale=None,
- lr_groups=None,
- embedding_lr=None,
- qlora=False,
- orpo_alpha=None,
- lisa_n_layers=None,
- lisa_step_interval=None,
- lisa_layers_attribute=None,
- curriculum_sampling=None,
- alternate_lr_scheduler_type=None,
- chat_template=None,
- kd_ce_alpha=None,
- kd_alpha=1.0,
- kd_temperature=1.0,
- kd_zscore_base_temp=None,
- kd_top_k_before_softmax=None,
- adam_beta3=None,
- adam_epsilon2=None,
- image_size=None,
- image_resize_algorithm=None,
- simpo_gamma=None,
-)core.training_args.AxolotlCPOConfig(simpo_gamma=None)CPO config for CPO training
core.training_args.AxolotlKTOConfig(
- model_type=None,
- lr_quadratic_warmup=False,
- pretraining=False,
- sample_packing=False,
- sample_packing_sequentially=False,
- multipack_real_batches=False,
- eval_sample_packing=None,
- sample_packing_efficiency=1.0,
- sample_packing_bin_size=200,
- sample_packing_group_size=100000,
- max_seq_length=2048,
- dataset_num_proc=None,
- relora_steps=None,
- relora_warmup_steps=None,
- relora_anneal_steps=None,
- relora_prune_ratio=0.9,
- bench_split='eval',
- bench_dataset='pharaouk/dharma-1/dharma_1_mini.json',
- do_bench_eval=False,
- do_causal_lm_eval=False,
- max_bench_samples=None,
- bench_source_max_len=2048,
- dataloader_prefetch_factor=None,
- cosine_min_lr_ratio=None,
- cosine_constant_lr_ratio=None,
- loraplus_lr_ratio=None,
- loraplus_lr_embedding=1e-06,
- embedding_lr_scale=None,
- lr_groups=None,
- embedding_lr=None,
- qlora=False,
- orpo_alpha=None,
- lisa_n_layers=None,
- lisa_step_interval=None,
- lisa_layers_attribute=None,
- curriculum_sampling=None,
- alternate_lr_scheduler_type=None,
- chat_template=None,
- kd_ce_alpha=None,
- kd_alpha=1.0,
- kd_temperature=1.0,
- kd_zscore_base_temp=None,
- kd_top_k_before_softmax=None,
- adam_beta3=None,
- adam_epsilon2=None,
- image_size=None,
- image_resize_algorithm=None,
-)core.training_args.AxolotlKTOConfig()KTO config for KTO training
core.training_args.AxolotlORPOConfig(
- model_type=None,
- lr_quadratic_warmup=False,
- pretraining=False,
- sample_packing=False,
- sample_packing_sequentially=False,
- multipack_real_batches=False,
- eval_sample_packing=None,
- sample_packing_efficiency=1.0,
- sample_packing_bin_size=200,
- sample_packing_group_size=100000,
- max_seq_length=2048,
- dataset_num_proc=None,
- relora_steps=None,
- relora_warmup_steps=None,
- relora_anneal_steps=None,
- relora_prune_ratio=0.9,
- bench_split='eval',
- bench_dataset='pharaouk/dharma-1/dharma_1_mini.json',
- do_bench_eval=False,
- do_causal_lm_eval=False,
- max_bench_samples=None,
- bench_source_max_len=2048,
- dataloader_prefetch_factor=None,
- cosine_min_lr_ratio=None,
- cosine_constant_lr_ratio=None,
- loraplus_lr_ratio=None,
- loraplus_lr_embedding=1e-06,
- embedding_lr_scale=None,
- lr_groups=None,
- embedding_lr=None,
- qlora=False,
- orpo_alpha=None,
- lisa_n_layers=None,
- lisa_step_interval=None,
- lisa_layers_attribute=None,
- curriculum_sampling=None,
- alternate_lr_scheduler_type=None,
- chat_template=None,
- kd_ce_alpha=None,
- kd_alpha=1.0,
- kd_temperature=1.0,
- kd_zscore_base_temp=None,
- kd_top_k_before_softmax=None,
- adam_beta3=None,
- adam_epsilon2=None,
- image_size=None,
- image_resize_algorithm=None,
-)core.training_args.AxolotlORPOConfig()ORPO config for ORPO training
core.training_args.AxolotlPRMConfig(
- model_type=None,
- lr_quadratic_warmup=False,
- pretraining=False,
- sample_packing=False,
- sample_packing_sequentially=False,
- multipack_real_batches=False,
- eval_sample_packing=None,
- sample_packing_efficiency=1.0,
- sample_packing_bin_size=200,
- sample_packing_group_size=100000,
- max_seq_length=2048,
- dataset_num_proc=None,
- relora_steps=None,
- relora_warmup_steps=None,
- relora_anneal_steps=None,
- relora_prune_ratio=0.9,
- bench_split='eval',
- bench_dataset='pharaouk/dharma-1/dharma_1_mini.json',
- do_bench_eval=False,
- do_causal_lm_eval=False,
- max_bench_samples=None,
- bench_source_max_len=2048,
- dataloader_prefetch_factor=None,
- cosine_min_lr_ratio=None,
- cosine_constant_lr_ratio=None,
- loraplus_lr_ratio=None,
- loraplus_lr_embedding=1e-06,
- embedding_lr_scale=None,
- lr_groups=None,
- embedding_lr=None,
- qlora=False,
- orpo_alpha=None,
- lisa_n_layers=None,
- lisa_step_interval=None,
- lisa_layers_attribute=None,
- curriculum_sampling=None,
- alternate_lr_scheduler_type=None,
- chat_template=None,
- kd_ce_alpha=None,
- kd_alpha=1.0,
- kd_temperature=1.0,
- kd_zscore_base_temp=None,
- kd_top_k_before_softmax=None,
- adam_beta3=None,
- adam_epsilon2=None,
- image_size=None,
- image_resize_algorithm=None,
-)core.training_args.AxolotlPRMConfig()PRM config for PRM training
core.training_args.AxolotlRewardConfig(
- model_type=None,
- lr_quadratic_warmup=False,
- pretraining=False,
- sample_packing=False,
- sample_packing_sequentially=False,
- multipack_real_batches=False,
- eval_sample_packing=None,
- sample_packing_efficiency=1.0,
- sample_packing_bin_size=200,
- sample_packing_group_size=100000,
- max_seq_length=2048,
- dataset_num_proc=None,
- relora_steps=None,
- relora_warmup_steps=None,
- relora_anneal_steps=None,
- relora_prune_ratio=0.9,
- bench_split='eval',
- bench_dataset='pharaouk/dharma-1/dharma_1_mini.json',
- do_bench_eval=False,
- do_causal_lm_eval=False,
- max_bench_samples=None,
- bench_source_max_len=2048,
- dataloader_prefetch_factor=None,
- cosine_min_lr_ratio=None,
- cosine_constant_lr_ratio=None,
- loraplus_lr_ratio=None,
- loraplus_lr_embedding=1e-06,
- embedding_lr_scale=None,
- lr_groups=None,
- embedding_lr=None,
- qlora=False,
- orpo_alpha=None,
- lisa_n_layers=None,
- lisa_step_interval=None,
- lisa_layers_attribute=None,
- curriculum_sampling=None,
- alternate_lr_scheduler_type=None,
- chat_template=None,
- kd_ce_alpha=None,
- kd_alpha=1.0,
- kd_temperature=1.0,
- kd_zscore_base_temp=None,
- kd_top_k_before_softmax=None,
- adam_beta3=None,
- adam_epsilon2=None,
- image_size=None,
- image_resize_algorithm=None,
-)core.training_args.AxolotlRewardConfig()Reward config for Reward training
core.training_args.AxolotlTrainingArguments(
- model_type=None,
- lr_quadratic_warmup=False,
- pretraining=False,
- sample_packing=False,
- sample_packing_sequentially=False,
- multipack_real_batches=False,
- eval_sample_packing=None,
- sample_packing_efficiency=1.0,
- sample_packing_bin_size=200,
- sample_packing_group_size=100000,
- max_seq_length=2048,
- dataset_num_proc=None,
- relora_steps=None,
- relora_warmup_steps=None,
- relora_anneal_steps=None,
- relora_prune_ratio=0.9,
- bench_split='eval',
- bench_dataset='pharaouk/dharma-1/dharma_1_mini.json',
- do_bench_eval=False,
- do_causal_lm_eval=False,
- max_bench_samples=None,
- bench_source_max_len=2048,
- dataloader_prefetch_factor=None,
- cosine_min_lr_ratio=None,
- cosine_constant_lr_ratio=None,
- loraplus_lr_ratio=None,
- loraplus_lr_embedding=1e-06,
- embedding_lr_scale=None,
- lr_groups=None,
- embedding_lr=None,
- qlora=False,
- orpo_alpha=None,
- lisa_n_layers=None,
- lisa_step_interval=None,
- lisa_layers_attribute=None,
- curriculum_sampling=None,
- alternate_lr_scheduler_type=None,
- chat_template=None,
- kd_ce_alpha=None,
- kd_alpha=1.0,
- kd_temperature=1.0,
- kd_zscore_base_temp=None,
- kd_top_k_before_softmax=None,
- adam_beta3=None,
- adam_epsilon2=None,
- image_size=None,
- image_resize_algorithm=None,
-)core.training_args.AxolotlTrainingArguments()Training arguments for Causal trainer
This code is duplicated due to HF TrainingArguments not setting output_dir with a default value so it can’t be used as a mixin.
-core.training_args.AxolotlTrainingMixins(
- model_type=None,
- lr_quadratic_warmup=False,
- pretraining=False,
- sample_packing=False,
- sample_packing_sequentially=False,
- multipack_real_batches=False,
- eval_sample_packing=None,
- sample_packing_efficiency=1.0,
- sample_packing_bin_size=200,
- sample_packing_group_size=100000,
- max_seq_length=2048,
- dataset_num_proc=None,
- relora_steps=None,
- relora_warmup_steps=None,
- relora_anneal_steps=None,
- relora_prune_ratio=0.9,
- bench_split='eval',
- bench_dataset='pharaouk/dharma-1/dharma_1_mini.json',
- do_bench_eval=False,
- do_causal_lm_eval=False,
- max_bench_samples=None,
- bench_source_max_len=2048,
- dataloader_prefetch_factor=None,
- cosine_min_lr_ratio=None,
- cosine_constant_lr_ratio=None,
- loraplus_lr_ratio=None,
- loraplus_lr_embedding=1e-06,
- embedding_lr_scale=None,
- lr_groups=None,
- embedding_lr=None,
- qlora=False,
- orpo_alpha=None,
- lisa_n_layers=None,
- lisa_step_interval=None,
- lisa_layers_attribute=None,
- curriculum_sampling=None,
- alternate_lr_scheduler_type=None,
- chat_template=None,
- kd_ce_alpha=None,
- kd_alpha=1.0,
- kd_temperature=1.0,
- kd_zscore_base_temp=None,
- kd_top_k_before_softmax=None,
- adam_beta3=None,
- adam_epsilon2=None,
- image_size=None,
- image_resize_algorithm=None,
-)Mixin class for the Axolotl training args.
integrations.base.BasePlugin.get_collator_cls_and_kwargs(cfg, is_eval=False)Returns a custom class for the collator.
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
| cfg | +DictDefault | +The global axolotl configuration. | +required | +
| is_eval | +bool | +Whether this is an eval split. | +False |
+
| Name | +Type | +Description | +
|---|---|---|
| class | ++ | The class for the collator. | +
integrations.base.BasePlugin.get_input_args()integrations.base.BasePlugin.get_input_args()Returns a pydantic model for the plugin’s input arguments.
integrations.base.BasePlugin.get_trainer_cls(cfg)integrations.base.BasePlugin.get_trainer_cls(cfg)Returns a custom class for the trainer.
-integrations.base.BasePlugin.get_training_args(cfg)Returns custom training arguments to set on TrainingArgs.
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
| cfg | +DictDefault | +The global axolotl configuration. | +required | +
| Name | +Type | +Description | +
|---|---|---|
| object | ++ | dict containing the training arguments. | +
integrations.base.BasePlugin.get_training_args_mixin()Returns a dataclass model for the plugin’s training arguments.
+integrations.base.BasePlugin.load_datasets(cfg, preprocess=False)integrations.base.BasePlugin.load_datasets(cfg, preprocess=False)Loads and preprocesses the dataset for training.
-| Name | -Type | -Description | -Default | -
|---|---|---|---|
| cfg | -DictDefault | -The configuration for the plugin. | -required | -
| model | -PreTrainedModel | PeftModel | -The loaded model. | -required | -
integrations.base.BasePlugin.post_model_build(cfg, model)Performs actions after the model is built/loaded, but before any adapters are applied.
-| Name | -Type | -Description | -Default | -
|---|---|---|---|
| cfg | -DictDefault | -The configuration for the plugin. | -required | -
integrations.base.BasePlugin.post_model_load(cfg, model)Performs actions after the model is loaded.
integrations.base.BasePlugin.post_train(cfg, model)Performs actions after training is complete.
+integrations.base.BasePlugin.post_model_build(cfg, model)Performs actions after the model is built/loaded, but before any adapters are applied.
| Name | +Type | +Description | +Default | +
|---|---|---|---|
| cfg | +DictDefault | +The configuration for the plugin. | +required | +
integrations.base.BasePlugin.post_model_load(cfg, model)Performs actions after the model is loaded.
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
| cfg | +DictDefault | +The configuration for the plugin. | +required | +
| model | +PreTrainedModel | PeftModel | +The loaded model. | +required | +
integrations.base.BasePlugin.post_train(cfg, model)Performs actions after training is complete.
+Calls the create_optimizer method of all registered plugins and returns |
|
| get_collator_cls_and_kwargs | +Calls the get_collator_cls_and_kwargs method of all registered plugins and returns the first non-None collator class. | +
| get_input_args | Returns a list of Pydantic classes for all registered plugins’ input arguments.’ |
| get_instance | Returns the singleton instance of PluginManager. If the instance doesn’t |
| get_trainer_cls | Calls the get_trainer_cls method of all registered plugins and returns the |
| get_training_args | +Calls the get_training_args method of all registered plugins and returns the combined training arguments. | +
| get_training_args_mixin | +Returns a list of dataclasses for all registered plugins’ training args mixins’ | +
| load_datasets | Calls the load_datasets method of each registered plugin. |
| post_lora_load | Calls the post_lora_load method of all registered plugins. |
| post_model_build | Calls the post_model_build method of all registered plugins after the |
| post_model_load | Calls the post_model_load method of all registered plugins after the model |
| post_train | Calls the post_train method of all registered plugins. |
| post_train_unload | Calls the post_train_unload method of all registered plugins. |
| post_trainer_create | Calls the post_trainer_create method of all registered plugins. |
| pre_lora_load | Calls the pre_lora_load method of all registered plugins. |
| pre_model_load | Calls the pre_model_load method of all registered plugins. |
| register | Registers a new plugin by its name. |
integrations.base.PluginManager.add_callbacks_post_trainer(cfg, trainer)integrations.base.PluginManager.add_callbacks_post_trainer(cfg, trainer)Calls the add_callbacks_post_trainer method of all registered plugins.
integrations.base.PluginManager.get_collator_cls_and_kwargs(cfg, is_eval=False)Calls the get_collator_cls_and_kwargs method of all registered plugins and returns the first non-None collator class.
+Parameters: +cfg (dict): The configuration for the plugins. +is_eval (bool): Whether this is an eval split.
+Returns: +object: The collator class, or None if none was found.
+integrations.base.PluginManager.get_input_args()integrations.base.PluginManager.get_input_args()Returns a list of Pydantic classes for all registered plugins’ input arguments.’
-integrations.base.PluginManager.get_training_args(cfg)Calls the get_training_args method of all registered plugins and returns the combined training arguments.
+Parameters: +cfg (dict): The configuration for the plugins.
+Returns: +object: The training arguments
+integrations.base.PluginManager.get_training_args_mixin()Returns a list of dataclasses for all registered plugins’ training args mixins’
+Returns: +list[str]: A list of dataclsses
+integrations.base.PluginManager.load_datasets(cfg, preprocess=False)integrations.base.PluginManager.load_datasets(cfg, preprocess=False)Calls the load_datasets method of each registered plugin.
-| Name | -Type | -Description | -Default | -
|---|---|---|---|
| cfg | -DictDefault | -The configuration for the plugins. | -required | -
| model | -PreTrainedModel | PeftModel | -The loaded model. | -required | -
integrations.base.PluginManager.post_model_build(cfg, model)Calls the post_model_build method of all registered plugins after the
-model has been built / loaded, but before any adapters have been applied.
| Name | -Type | -Description | -Default | -
|---|---|---|---|
| cfg | -DictDefault | -The configuration for the plugins. | -required | -
| model | -PreTrainedModel | -The loaded model. | -required | -
integrations.base.PluginManager.post_model_load(cfg, model)Calls the post_model_load method of all registered plugins after the model
-has been loaded inclusive of any adapters.
integrations.base.PluginManager.post_train(cfg, model)Calls the post_train method of all registered plugins.
+integrations.base.PluginManager.post_model_build(cfg, model)Calls the post_model_build method of all registered plugins after the
+model has been built / loaded, but before any adapters have been applied.
| Name | +Type | +Description | +Default | +
|---|---|---|---|
| cfg | +DictDefault | +The configuration for the plugins. | +required | +
| model | +PreTrainedModel | +The loaded model. | +required | +
integrations.base.PluginManager.post_model_load(cfg, model)Calls the post_model_load method of all registered plugins after the model
+has been loaded inclusive of any adapters.
| Name | +Type | +Description | +Default | +
|---|---|---|---|
| cfg | +DictDefault | +The configuration for the plugins. | +required | +
| model | +PreTrainedModel | PeftModel | +The loaded model. | +required | +
integrations.base.PluginManager.post_train(cfg, model)Calls the post_train method of all registered plugins.
+integrations.base.load_plugin(plugin_name)integrations.base.load_plugin(plugin_name)Loads a plugin based on the given plugin name.
The plugin name should be in the format “module_name.class_name”. This function splits the plugin name into module and class, imports the module, retrieves the class from the module, and creates an instance of the class.
-integrations.kd.trainer.AxolotlKDTrainer(
- *_args,
- bench_data_collator=None,
- eval_data_collator=None,
- dataset_tags=None,
- **kwargs,
-)integrations.kd.trainer.AxolotlKDTrainer(*args, **kwargs)Custom trainer subclass for Knowledge Distillation (KD)
| - | tuple[HFRLTrainerBuilder | HFCausalTrainerBuilder, PeftModel | PreTrainedModel, PreTrainedTokenizer, PeftConfig | None, ProcessorMixin | None] | +tuple['HFRLTrainerBuilder' | 'HFCausalTrainerBuilder', PeftModel | PreTrainedModel, PreTrainedTokenizer, PeftConfig | None, ProcessorMixin | None] | Tuple of: - Trainer (Causal or RLHF) - Model - Tokenizer - PEFT config - Processor |