easydel.trainers.group_relative_policy_optimization.grpo_config

class easydel.trainers.group_relative_policy_optimization.grpo_config.GRPOConfig(auto_shard_states: bool = True, aux_loss_enabled: bool = False, backend: str | None = None, clip_grad: float | None = None, custom_scheduler: tp.Callable[[int], tp.Any] | None = None, dataloader_num_workers: int | None = 0, dataloader_pin_memory: bool | None = False, do_eval: bool = False, do_last_save: bool = True, do_train: bool = True, eval_batch_size: int | None = None, evaluation_steps: int | None = None, extra_optimizer_kwargs: dict = <factory>, frozen_parameters: str | None = None, grain_shard_index: int | None = None, grain_shard_count: int | None = None, gradient_accumulation_steps: int = 1, ids_to_pop_from_dataset: list[str] | None = <factory>, is_fine_tuning: bool = True, init_tx: bool = True, jax_distributed_config: dict | None = None, learning_rate: float = 1e-06, learning_rate_end: float | None = None, log_all_workers: bool = False, log_grad_norms: bool = True, report_metrics: bool = True, log_steps: int = 10, loss_config: LossConfig | None = None, low_mem_usage: bool = True, max_evaluation_steps: int | None = None, max_sequence_length: int | None = 4096, max_training_steps: int | None = None, per_epoch_training_steps: int | None = None, per_epoch_evaluation_steps: int | None = None, model_name: str | None = None, model_parameters: dict | None = None, metrics_to_show_in_rich_pbar: list[str] | None = None, generation_top_p: float | None = None, generation_top_k: int | None = None, generation_temperature: float | None = None, generation_do_sample: bool | None = None, generation_num_return_sequences: int | None = None, generation_max_new_tokens: int | None = None, generation_shard_inputs: bool = True, generation_interval: int | None = None, generation_prompts: list[str | dict[str, tp.Any]] = <factory>, generation_use_train_prompts: bool = False, generation_num_prompts: int = 1, generation_dataset_prompt_field: str | None = 'prompt', generation_extra_kwargs: dict[str, tp.Any] | None = None, generation_config_overrides: dict[str, tp.Any] | None = None, generation_seed: int | None = None, generation_preview_print: bool = False, generation_log_to_wandb: bool = True, use_esurge_generation: bool = True, esurge_use_tqdm: bool = True, esurge_hbm_utilization: float | None = 0.45, esurge_max_num_seqs: int | None = None, esurge_min_input_pad: int | None = None, esurge_page_size: int | None = 32, esurge_silent_mode: bool = True, num_train_epochs: int = 10, offload_dataset: bool = False, offload_device_type: str = 'cpu', offload_device_index: int = 0, optimizer: AVAILABLE_OPTIMIZERS = EasyDeLOptimizers.ADAMW, performance_mode: bool = False, pruning_module: tp.Any = None, process_zero_is_admin: bool = True, progress_bar_type: tp.Literal['tqdm', 'rich', 'json'] = 'tqdm', remove_ckpt_after_load: bool = False, remove_unused_columns: bool | None = False, report_steps: int = 5, save_interval_minutes: float | None = None, save_directory: str = 'EasyDeL-Checkpoints', save_optimizer_state: bool = True, save_steps: int | None = None, save_total_limit: int | None = None, scheduler: AVAILABLE_SCHEDULERS = EasyDeLSchedulers.NONE, shuffle_seed_train: int = 64871, sparsify_module: bool = False, sparse_module_type: AVAILABLE_SPARSE_MODULE_TYPES = 'bcoo', state_apply_fn_kwarguments_to_model: dict | None = None, step_partition_spec: PartitionSpec = PartitionSpec(('dp', 'fsdp'), 'sp'), step_start_point: int | None = None, resume_if_possible: bool = True, shuffle_train_dataset: bool = True, total_batch_size: int = 32, training_time_limit: str | None = None, train_on_inputs: bool = True, trainer_prefix: str | None = 'grpotrainer', truncation_mode: tp.Literal['keep_end', 'keep_start'] = 'keep_end', tx_mu_dtype: jnp.dtype | None = None, track_memory: bool | float = False, use_data_collactor: bool = True, use_grain: bool = True, use_wandb: bool = True, verbose: bool = True, wandb_entity: str | None = None, wandb_name: str | None = None, warmup_steps: int = 0, weight_decay: float = 0.01, weight_distribution_pattern: str = '.*', weight_distribution_log_steps: int = 50, _can_log_metrics: bool | None = None, _im_a_hidden_checkpoint_manager: Checkpointer | None = None, max_prompt_length: int = 512, max_completion_length: int = 256, dataset_num_proc: int | None = None, beta: float = 0.04, epsilon: float = 0.2, epsilon_high: float | None = None, delta: float | None = None, sync_ref_model: bool = False, ref_model_mixup_alpha: float = 0.9, ref_model_sync_steps: int = 64, num_iterations: int = 1, loss_type: str = 'dapo', importance_sampling_level: str = 'token', reward_weights: list[float] | None = None, scale_rewards: str | bool = 'group', tools: list[typing.Union[dict, typing.Callable]] | None = None, skip_apply_chat_template: bool = False, num_return_sequences: int = 4, num_generations: int | None = None, temperature: float = 1.0, top_p: float = 1.0, top_k: int | None = None, min_p: float | None = None, repetition_penalty: float = 1.0, generation_kwargs: dict | None = None, chat_template_kwargs: dict | None = None, mask_truncated_completions: bool = False, top_entropy_quantile: float = 1.0)

Bases: TrainingArguments

Configuration class for Group Relative Policy Optimization training.

GRPO is an efficient RLHF algorithm that optimizes policies using group-based relative comparisons of rewards. It provides better training stability compared to standard PPO by normalizing rewards within groups of samples.

This configuration extends TrainingArguments with GRPO-specific parameters for controlling the policy optimization process, reward computation, and generation sampling strategies.

Key concepts: - Group-based normalization: Rewards are normalized within groups to reduce variance - KL regularization: Prevents the policy from deviating too far from reference - Reference model syncing: Optionally updates reference model during training

beta: float = 0.04

chat_template_kwargs: dict | None = None

dataset_num_proc: int | None = None

delta: float | None = None

epsilon: float = 0.2

epsilon_high: float | None = None

classmethod from_dict(data: dict[str, Any]) → T

Deserializes a dictionary into a PyTree object.

classmethod from_json(json_str: str) → T

Deserializes a JSON string into a PyTree object.

generation_kwargs: dict | None = None

importance_sampling_level: str = 'token'

learning_rate: float = 1e-06

loss_type: str = 'dapo'

mask_truncated_completions: bool = False

max_completion_length: int = 256

max_prompt_length: int = 512

min_p: float | None = None

num_generations: int | None = None

num_iterations: int = 1

num_return_sequences: int = 4

ref_model_mixup_alpha: float = 0.9

ref_model_sync_steps: int = 64

remove_unused_columns: bool | None = False

repetition_penalty: float = 1.0

replace(**kwargs)

Creates a new instance with specified fields replaced.

reward_weights: list[float] | None = None

scale_rewards: str | bool = 'group'

skip_apply_chat_template: bool = False

sync_ref_model: bool = False

temperature: float = 1.0

to_dict() → dict[str, Any]

Serializes the PyTree object to a dictionary.

to_json(**kwargs) → str

Serializes the PyTree object to a JSON string.

tools: list[Union[dict, Callable]] | None = None

top_entropy_quantile: float = 1.0

top_k: int | None = None

top_p: float = 1.0

trainer_prefix: str | None = 'grpotrainer'