Functions — github.com/OpenLMLab/MOSS-RLHF

CodeHub MCPcopy Create free account

hub / github.com/OpenLMLab/MOSS-RLHF / functions

Functions158 in github.com/OpenLMLab/MOSS-RLHF

⨍ Functions 158◇ Types & classes 20

↓ 36 callersMethodcreate_metric

(self, metric_name: str, metric_obj: Metric)

↓ 33 callersMethodrecord_metric

(self, metric_name: str, val: Any)

↓ 20 callersFunctionprint_rank_0

(info, only_on_cuda0=False)

↓ 15 callersFunctionpad_sequences

Padding sequence to the same length

↓ 10 callersFunctionsynchronize_if_distributed

↓ 9 callersMethodrecord_metric_many

(self, metric_name: str, vals: List[Any], counts: Optional[List[int]] = None)

↓ 8 callersMethodstate_dict

ppo/ppo_trainer.py:22

↓ 6 callersMethodall_gather_metrics

↓ 6 callersFunctionto_cuda

↓ 5 callersMethodcompute

(self, val: Any)

↓ 5 callersMethodeval

ppo/ppo_trainer.py:42

↓ 5 callersFunctionget_tokenizer

ppo/ppo_datahelper.py:21

↓ 5 callersMethodreset

↓ 5 callersMethodtrain

ppo/ppo_trainer.py:613

↓ 4 callersMethodbatch_generator

ppo/ppo_datahelper.py:76

↓ 4 callersMethoddisplay

(self, global_step, data_size = None, gathered_metrics: Dict[str, float] = None)

↓ 4 callersFunctionget_tokenizer

rm/reward_datahelper.py:20

↓ 4 callersMethodwrite_tensorboard

(self, global_step, gathered_metrics: Dict[str, float] = None)

↓ 3 callersMethodformat

(self, sample: Tuple[List[str], str])

ppo/ppo_datahelper.py:314

↓ 3 callersFunctionget_assistant_prompt

ppo/ppo_datahelper.py:17

↓ 3 callersFunctionget_human_prompt

ppo/ppo_datahelper.py:13

↓ 3 callersFunctionlogprobs_from_logits

See: https://github.com/pytorch/pytorch/issues/563#issuecomment-330103591

↓ 3 callersMethodpolicy_model_forward

(self, inputs, **kwargs)

ppo/ppo_trainer.py:227

↓ 3 callersMethodval

↓ 2 callersMethod_build_prompt

(self, context: List[str], dialog_sep='\n')

rm/reward_datahelper.py:106

↓ 2 callersMethod_criterion

(self, model_output: torch.Tensor, batch: Dict[str, Any], return_output=False, training=True)

rm/reward_trainer.py:134

↓ 2 callersMethod_get_metric_obj

(self, training=True)

rm/reward_trainer.py:248

↓ 2 callersMethod_on_stop_train

rm/reward_trainer.py:297

↓ 2 callersMethod_record_batch_info

(self, batch, mode='train')

rm/reward_trainer.py:254

↓ 2 callersMethod_run_forward

(self, batch: Dict[str, Any], **kwargs)

rm/reward_trainer.py:271

↓ 2 callersMethodadd

↓ 2 callersMethodbuild_dataloader

(self, mode='train')

rm/reward_trainer.py:194

↓ 2 callersMethodbuild_metrics

(self, mode='train')

rm/reward_trainer.py:84

↓ 2 callersMethodbuild_metrics

(self, mode='train')

ppo/ppo_trainer.py:135

↓ 2 callersMethodconcat_context_and_response

(self, context: List[List[int]], responses: List[List[Tuple[float, List[int]]]])

ppo/ppo_trainer.py:239

↓ 2 callersMethodevaluate

(self, datatype='valid', **kwargs)

rm/reward_trainer.py:363

↓ 2 callersMethodevaluate

(self, datatype='valid', **kwargs)

ppo/ppo_trainer.py:551

↓ 2 callersMethodflush

↓ 2 callersMethodforward

(self, decoder_input, incr_state=None)

train_ppo.py:21

↓ 2 callersMethodgenerate

Generate response

train_ppo.py:41

↓ 2 callersFunctionget_assistant_prompt

rm/reward_datahelper.py:14

↓ 2 callersFunctionget_global_statistics

Computes element-wise mean and variance of the tensor across processes https://github.com/microsoft/LMOps/blob/cde1fb1ef4608a7ac5bf00675fa3e9

↓ 2 callersFunctionget_human_prompt

rm/reward_datahelper.py:10

↓ 2 callersFunctionget_model_prompt

(context: List[str], eos_token="</s>", opt=None)

ppo/ppo_datahelper.py:43

↓ 2 callersMethodget_parms

(self, model, submodel_name, lr, eps)

ppo/ppo_trainer.py:178

↓ 2 callersFunctionget_special_prompt

ppo/ppo_datahelper.py:40

↓ 2 callersMethodmany

(self, vals: List[Any])

↓ 2 callersMethodref_model_forward

(self, inputs, **kwargs)

ppo/ppo_trainer.py:230

↓ 2 callersMethodreward_model_forward

(self, inputs, **kwargs)

ppo/ppo_trainer.py:224

↓ 2 callersFunctionsetup_accelerator

↓ 2 callersMethodstrip_pad_token_id

(self, seq: List[int])

ppo/ppo_trainer.py:212

↓ 2 callersFunctionsynchronize_forward_on_stage3

(done: bool, fake_forward_fn: Callable, **kwargs)

↓ 2 callersMethodupdate

Updates running moments from batch's moments computed across ranks

↓ 1 callersMethodRLHF_model_forward

(self, batch: Dict[str, Any], **kwargs)

ppo/ppo_trainer.py:236

↓ 1 callersMethod__add__

↓ 1 callersMethod__init__

(self, config, opt, tokenizer)

train_ppo.py:16

↓ 1 callersMethod__init__

ppo/ppo_datahelper.py:56

↓ 1 callersMethod__init__

ppo/ppo_trainer.py:17

↓ 1 callersMethod_batchify

(self, batch_samples: List[Dict[str, Any]])

rm/reward_datahelper.py:157

↓ 1 callersMethod_encode_sample

(self, sample: Tuple[List[str], str, str])

rm/reward_datahelper.py:120

↓ 1 callersMethod_lm_loss

(self, scores:torch.Tensor, preds:torch.Tensor, labels:torch.LongTensor, training=True)

rm/reward_trainer.py:111

↓ 1 callersMethod_load_data

(self, dpath: str)

rm/reward_datahelper.py:83

↓ 1 callersMethod_save_checkpoint

(self, is_best: bool, total_steps: int)

rm/reward_trainer.py:169

↓ 1 callersMethod_train_step

(self, batch: Dict[str, Any], **kwargs)

rm/reward_trainer.py:279

↓ 1 callersMethodadd

↓ 1 callersMethodbatch_generator

rm/reward_datahelper.py:189

↓ 1 callersMethodbatchify

(self, batch_samples: List[Dict[str, Any]])

ppo/ppo_datahelper.py:337

↓ 1 callersMethodbuild_optimizer

rm/reward_trainer.py:223

↓ 1 callersMethodbuild_optimizer

ppo/ppo_trainer.py:203

↓ 1 callersMethodbuild_scheduler

rm/reward_trainer.py:241

↓ 1 callersMethodcriterion

(self, model_output: Tuple[torch.Tensor, ...], batch: Dict[str, Any], return_output=False, training=True)

ppo/ppo_trainer.py:391

↓ 1 callersMethodcritic_model_forward

(self, inputs, **kwargs)

ppo/ppo_trainer.py:233

↓ 1 callersMethodfinal_generator

ppo/ppo_datahelper.py:92

↓ 1 callersMethodget_advantages_and_returns

Copied from TRLX: https://github.com/CarperAI/trlx/blob/main/trlx/models/modeling_ppo.py

ppo/ppo_datahelper.py:192

↓ 1 callersFunctionget_category_distribution_entropy

Compute category distribution entropy

↓ 1 callersFunctionget_eval_ds_config

(offload=None, stage=3)

↓ 1 callersMethodget_optimizer_grouped_parameters

(self, model, weight_decay)

rm/reward_trainer.py:203

↓ 1 callersFunctionget_separate_prompt

rm/reward_datahelper.py:17

↓ 1 callersMethodinvsqrt_scheduler

(self, warmup_steps)

rm/reward_trainer.py:231

↓ 1 callersMethodinvsqrt_scheduler

(self, warmup_steps)

ppo/ppo_trainer.py:168

↓ 1 callersMethodload_data

(self, file_path: str)

ppo/ppo_datahelper.py:130

↓ 1 callersMethodload_data

(self, file_path: str)

ppo/ppo_datahelper.py:300

↓ 1 callersFunctionmain

train_ppo.py:137

↓ 1 callersFunctionmain

↓ 1 callersMethodmake_experiences

ppo/ppo_trainer.py:285

↓ 1 callersFunctionparse_args

config_ppo.py:3

↓ 1 callersFunctionparse_args

↓ 1 callersMethodpretrain_dataset_class

ppo/ppo_trainer.py:215

↓ 1 callersMethodsample_generator

rm/reward_datahelper.py:174

↓ 1 callersMethodsample_generator

ppo/ppo_datahelper.py:62

↓ 1 callersMethodsave_checkpoint

(self, is_best: bool, total_steps: int)

ppo/ppo_trainer.py:258

↓ 1 callersFunctiontop_p_logits

Filter a distribution of logits using nucleus (top-p) filtering https://github.com/OpenLMLab/MOSS/blob/e088f438d1a95d424c6dffef0d73134ebe62cb

↓ 1 callersMethodtrain

rm/reward_trainer.py:301

↓ 1 callersMethodtrain_step

(self, batch: Dict[str, Any], **kwargs)

ppo/ppo_trainer.py:527

↓ 1 callersFunctionwhiten

(self, other: 'MeanMetric')

(self, other: 'SumMetric')

next →1–100 of 158, ranked by callers