MCPcopy Create free account

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

Functions158 in github.com/OpenLMLab/MOSS-RLHF

Method__init__
(self, num=0, denom=0)
metric.py:31
Method__init__
(self, sum_=0)
metric.py:60
Method__init__
(self, val=0)
metric.py:80
Method__init__
(self, opt: Dict[str, Any], accelerator, mode='train')
metric.py:111
Method__init__
(self, config, opt, tokenizer)
train_ppo.py:114
Method__init__
Calculates the running mean and standard deviation of a data stream. Modified version of https://github.com/DLR-RM/stable-baselines3/
utils.py:89
Method__init__
(self, config, opt: Dict[str, Any], tokenizer, **kwargs)
train_rm.py:14
Method__init__
(self, opt, accelerator, mode: str = 'train', **kwargs)
rm/reward_datahelper.py:38
Method__init__
(self)
rm/reward_trainer.py:23
Method__init__
(self, opt, model: nn.Module, accelerator, eval_only=False)
rm/reward_trainer.py:43
Method__init__
(self, opt, accelerator, mode = 'train', **kwargs)
ppo/ppo_datahelper.py:103
Method__init__
(self, data, opt, accelerator, mode = 'train', **kwargs)
ppo/ppo_datahelper.py:176
Method__init__
(self, opt, accelerator, **kwargs)
ppo/ppo_datahelper.py:275
Method__init__
(self, policy_model, critic_model)
ppo/ppo_trainer.py:30
Method__init__
(self, opt, policy_model, ref_model, critic_model, reward_model, accelerator, **kwargs)
ppo/ppo_trainer.py:48
Method__iter__
(self)
rm/reward_datahelper.py:75
Method__iter__
(self)
ppo/ppo_datahelper.py:98
Method__len__
(self)
rm/reward_datahelper.py:72
Method__len__
(self)
ppo/ppo_datahelper.py:59
Method__radd__
(self, other)
metric.py:26
Method_invsqrt_lr
(step)
rm/reward_trainer.py:232
Method_invsqrt_lr
(step)
ppo/ppo_trainer.py:169
Method_invsqrt_lr_with_warmup
(step)
rm/reward_trainer.py:236
Method_invsqrt_lr_with_warmup
(step)
ppo/ppo_trainer.py:173
Method_run_fake_forward
(self)
rm/reward_trainer.py:106
Method_warmup_lr
(step)
rm/reward_trainer.py:234
Method_warmup_lr
(step)
ppo/ppo_trainer.py:171
Methodadd
(self, val: Any)
metric.py:35
Methodadd
(self, val)
metric.py:63
Methodbatch_generator
(self)
ppo/ppo_datahelper.py:166
Methodbatch_generator
(self)
ppo/ppo_datahelper.py:232
Methodbatch_generator
(self)
ppo/ppo_datahelper.py:353
Methodbatchify
(self, batch_samples: List[Dict[str, Any]])
ppo/ppo_datahelper.py:157
Methodbatchify
(self, batch_samples: List[Dict[str, Any]])
ppo/ppo_datahelper.py:237
Methodformat
(self, sample: List[str])
ppo/ppo_datahelper.py:139
Methodformat
(self, sample: Dict[str, Any])
ppo/ppo_datahelper.py:210
Methodforward
(self, decoder_input, only_last=True)
train_ppo.py:120
Methodforward
(self, decoder_input: torch.LongTensor, rank_all=False)
train_rm.py:26
Methodforward
(self, inputs, **kwargs)
ppo/ppo_trainer.py:35
Methodload_state_dict
(self, state_dict)
rm/reward_trainer.py:35
Functionmain
(task, **kwargs)
merge_weight_en.py:176
Functionmain
(task, **kwargs)
merge_weight_zh.py:132
Functionmake_diff
Make the weight diff. This function is given to present full transparency of how the weight diff was created. Run: python weight_dif
merge_weight_en.py:25
Functionmake_diff
Make the weight diff. This function is given to present full transparency of how the weight diff was created. Run: python weight_dif
merge_weight_zh.py:25
Methodmany
(self, vals: List[Any], denoms: Optional[List[int]] = None)
metric.py:39
Methodmany
(self, vals: List[Any])
metric.py:86
Functionrecover
Recover the original weights from the released weight diff. This function is given for you to run. Things to do before running this:
merge_weight_en.py:90
Functionrecover
Recover the original weights from the released weight diff. This function is given for you to run. Things to do before running this:
merge_weight_zh.py:69
Methodreset
(self)
metric.py:53
Methodreset
(self)
metric.py:72
Methodreset
(self)
metric.py:92
Methodreset
(self, no_reset = ['global_exs'])
metric.py:133
Methodstate_dict
(self)
rm/reward_trainer.py:28
Methodtrain
(self, mode=True)
ppo/ppo_trainer.py:38
Methodval
(self)
metric.py:48
Methodval
(self)
metric.py:69
Methodval
(self)
metric.py:89
Methodval
(self)
metric.py:99
← previous101–158 of 158, ranked by callers