Index

A

add_extra_args() (chatlearn.models.megatron_module.MegatronModule method) add_padding_config() (chatlearn.models.base_module.BaseModule method)

after_episode() (chatlearn.models.base_module.BaseModule method) args_dict (chatlearn.utils.arguments.ModelConfig attribute)

B

BaseModule (class in chatlearn.models.base_module) batch_generation (chatlearn.utils.arguments.ModelConfig attribute) BatchGenerationConfig (class in chatlearn.utils.arguments) before_episode() (chatlearn.models.base_module.BaseModule method)

bucket_size_mb_in_memory_manager (chatlearn.utils.arguments.RuntimeConfig attribute) build_dataset() (chatlearn.models.base_module.BaseModule method) build_grad_buffers() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method)

C

coalesce_param (chatlearn.utils.arguments.RuntimeConfig attribute) coalesced_buffer_mb (chatlearn.utils.arguments.RuntimeConfig attribute) colocation (chatlearn.utils.arguments.RuntimeConfig attribute) column_only_qkv (chatlearn.utils.arguments.LoraConfig attribute) concat_params_dict (chatlearn.models.base_module.BaseModule property)

concurrent_comm (chatlearn.utils.arguments.RuntimeConfig attribute) concurrent_setup (chatlearn.utils.arguments.RuntimeConfig attribute) config_dir (chatlearn.utils.arguments.ModelConfig attribute) consumed_samples (chatlearn.utils.arguments.RuntimeConfig attribute) cpu_per_process (chatlearn.utils.arguments.ModelConfig attribute) cpu_schedule_strategy (chatlearn.utils.arguments.RuntimeConfig attribute)

D

data_checkpoint_path (chatlearn.utils.arguments.RuntimeConfig attribute) data_path (chatlearn.utils.arguments.RuntimeConfig attribute)

debug (chatlearn.utils.arguments.RuntimeConfig attribute) destroy_collective_group() (chatlearn.models.base_module.BaseModule method) DPOEngine (class in chatlearn)

E

enable_lora (chatlearn.utils.arguments.LoraConfig attribute) enable_resume_training (chatlearn.utils.arguments.RuntimeConfig attribute) eval() (chatlearn.EvalEngine method) (chatlearn.Evaluator method) eval_episode_interval (chatlearn.utils.arguments.RuntimeConfig attribute)

eval_step() (chatlearn.models.base_module.BaseModule method) EvalEngine (class in chatlearn) Evaluator (class in chatlearn) excludes (chatlearn.utils.arguments.RuntimeEnvConfig attribute) exp_name (chatlearn.utils.arguments.RuntimeConfig attribute)

F

forward_step() (chatlearn.models.base_module.BaseModule method) free_grad_buffers (chatlearn.utils.arguments.ModelConfig attribute) free_grad_buffers() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method)

free_memory (chatlearn.utils.arguments.ModelConfig attribute) free_sync_collective_group (chatlearn.utils.arguments.RuntimeConfig attribute)

G

generation_batch_size (chatlearn.utils.arguments.ModelConfig attribute) (chatlearn.utils.arguments.RuntimeConfig attribute) get() (chatlearn.models.base_module.BaseModule method) (chatlearn.utils.arguments.RuntimeConfig method) (chatlearn.utils.arguments.RuntimeEnvConfig method) get_concat_params_dict() (chatlearn.models.base_module.BaseModule method)

get_data_parallel_rank() (chatlearn.models.base_module.BaseModule method) get_data_parallel_size() (chatlearn.models.base_module.BaseModule method) get_to_fix_act_ordering_dict() (chatlearn.models.base_module.BaseModule method) get_to_fix_qkv_ordering_dict() (chatlearn.models.base_module.BaseModule method) get_to_fix_qkv_ordering_func() (chatlearn.models.base_module.BaseModule method) gpu_per_process (chatlearn.utils.arguments.ModelConfig attribute)

I

init() (chatlearn.models.base_module.BaseModule method) is_colocate (chatlearn.models.base_module.BaseModule property)

is_last_rank() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.torch_module.TorchModule method) is_master_node() (chatlearn.models.base_module.BaseModule method)

L

load_data_checkpoint_iteration (chatlearn.utils.arguments.RuntimeConfig attribute) log_interval (chatlearn.utils.arguments.RuntimeConfig attribute) lora (chatlearn.utils.arguments.ModelConfig attribute) lora_dim (chatlearn.utils.arguments.LoraConfig attribute)

lora_dropout (chatlearn.utils.arguments.LoraConfig attribute) lora_layer (chatlearn.utils.arguments.LoraConfig attribute) lora_scaling (chatlearn.utils.arguments.LoraConfig attribute) LoraConfig (class in chatlearn.utils.arguments)

M

max_data_ckpt_nums (chatlearn.utils.arguments.RuntimeConfig attribute) max_relay_episode (chatlearn.utils.arguments.RuntimeConfig attribute) megatron_model() (chatlearn.models.megatron_module.MegatronModule method) MegatronModule (class in chatlearn.models.megatron_module) min_prompt_length (chatlearn.utils.arguments.BatchGenerationConfig attribute)

model_args (chatlearn.models.base_module.BaseModule property) model_config_file (chatlearn.utils.arguments.ModelConfig attribute) model_type (chatlearn.utils.arguments.ModelConfig attribute) ModelConfig (class in chatlearn.utils.arguments) module_args (chatlearn.models.base_module.BaseModule property)

N

nsys (chatlearn.utils.arguments.RuntimeConfig attribute) num_cpu (chatlearn.utils.arguments.ModelConfig attribute) num_device (chatlearn.utils.arguments.ModelConfig attribute)

num_episode (chatlearn.utils.arguments.RuntimeConfig attribute) num_gpu (chatlearn.utils.arguments.ModelConfig attribute) num_replica (chatlearn.utils.arguments.ModelConfig attribute) num_training_epoch (chatlearn.utils.arguments.RuntimeConfig attribute)

O

offload() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.torch_module.TorchModule method) offload_main_weights() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method) offload_optimizer_states (chatlearn.utils.arguments.ModelConfig attribute) offload_optimizer_states() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method) offload_weights (chatlearn.utils.arguments.ModelConfig attribute) offload_weights() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method)

OnlineDPOEngine (class in chatlearn) onload() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.torch_module.TorchModule method) onload_main_weights() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method) onload_optimizer_states() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method) onload_weights() (chatlearn.models.base_module.BaseModule method) (chatlearn.models.megatron_module.MegatronModule method) output_dir (chatlearn.utils.arguments.RuntimeConfig attribute)

P

param_sync_comm_type (chatlearn.utils.arguments.RuntimeConfig attribute) param_sync_max_workers (chatlearn.utils.arguments.RuntimeConfig attribute) parameter_sync_frequency (chatlearn.models.base_module.BaseModule property) part_module_name (chatlearn.utils.arguments.LoraConfig attribute) pip (chatlearn.utils.arguments.RuntimeEnvConfig attribute)

pipeline_model_parallel_size (chatlearn.utils.arguments.ModelConfig attribute) platform (chatlearn.utils.arguments.RuntimeEnvConfig attribute) profiler_dir (chatlearn.utils.arguments.RuntimeConfig attribute) put() (chatlearn.models.base_module.BaseModule method) py_modules (chatlearn.utils.arguments.RuntimeEnvConfig attribute)

R

ranking (chatlearn.utils.arguments.BatchGenerationConfig attribute) relay_episode_offset (chatlearn.utils.arguments.RuntimeConfig attribute) resume_training (chatlearn.models.base_module.BaseModule property)

RLHFEngine (class in chatlearn) runtime_args (chatlearn.models.base_module.BaseModule property) RuntimeConfig (class in chatlearn.utils.arguments) RuntimeEnvConfig (class in chatlearn.utils.arguments)

S

sample_per_episode (chatlearn.utils.arguments.RuntimeConfig attribute) save_checkpoint() (chatlearn.models.base_module.BaseModule method) save_episode_interval (chatlearn.utils.arguments.RuntimeConfig attribute) set_colocate() (chatlearn.models.base_module.BaseModule method) set_concat_params_dict() (chatlearn.models.base_module.BaseModule method) set_dataset() (chatlearn.EvalEngine method) (chatlearn.Evaluator method) set_post_process_func() (chatlearn.EvalEngine method) (chatlearn.Evaluator method)

set_src_parameter_model() (chatlearn.models.base_module.BaseModule method) set_to_fix_act_ordering_dict() (chatlearn.models.base_module.BaseModule method) set_to_fix_qkv_ordering_dict() (chatlearn.models.base_module.BaseModule method) set_to_fix_qkv_ordering_func() (chatlearn.models.base_module.BaseModule method) setup() (chatlearn.models.base_module.BaseModule method) src_parameter_model (chatlearn.models.base_module.BaseModule property) stream_data_loader_type (chatlearn.utils.arguments.RuntimeConfig attribute) sync_frequency (chatlearn.utils.arguments.ModelConfig attribute)

T

tensor_model_parallel_size (chatlearn.utils.arguments.ModelConfig attribute) to_fix_act_ordering_dict (chatlearn.models.base_module.BaseModule property) to_fix_qkv_ordering_dict (chatlearn.models.base_module.BaseModule property) to_fix_qkv_ordering_func (chatlearn.models.base_module.BaseModule property)

TorchModule (class in chatlearn.models.torch_module) train_global_batch_size (chatlearn.utils.arguments.RuntimeConfig attribute) train_micro_batch_size (chatlearn.utils.arguments.RuntimeConfig attribute) train_step() (chatlearn.models.base_module.BaseModule method) trainable (chatlearn.utils.arguments.ModelConfig attribute)

W

working_dir (chatlearn.utils.arguments.RuntimeEnvConfig attribute)

world_size (chatlearn.models.base_module.BaseModule property) (chatlearn.models.torch_module.TorchModule property)

Z

zero_size (chatlearn.utils.arguments.ModelConfig attribute)