MCPcopy Create free account

hub / github.com/Shredded-Pork/TempFlow-GRPO / functions

Functions220 in github.com/Shredded-Pork/TempFlow-GRPO

⨍ Functions 220◇ Types & classes 28
↓ 143 callersMethodto
(self, device: torch.device = None, dtype: torch.dtype = None)
flow_grpo/ema.py:58
↓ 21 callersMethodupdate
(self, prompts, rewards, exp=False)
flow_grpo/stat_tracking.py:12
↓ 19 callersFunctionsde_step_with_logprob
( self: FlowMatchEulerDiscreteScheduler, model_output: torch.FloatTensor, timestep: Union[float, t
flow_grpo/diffusers_patch/sd3_sde_with_logprob.py:17
↓ 10 callersMethodcopy_ema_to
(self, parameters: Iterable[torch.nn.Parameter], store_temp: bool = True)
flow_grpo/ema.py:65
↓ 10 callersMethodcopy_temp_to
(self, parameters: Iterable[torch.nn.Parameter])
flow_grpo/ema.py:73
↓ 10 callersMethodstep
(self, parameters: Iterable[torch.nn.Parameter], optimization_step)
flow_grpo/ema.py:40
↓ 9 callersFunctioncompressibility
()
config/dgx.py:7
↓ 7 callersMethodclear
(self)
flow_grpo/stat_tracking.py:41
↓ 7 callersMethodget_stats
(self)
flow_grpo/stat_tracking.py:36
↓ 6 callersFunctionfsdp_wrapper
(model, fsdp_config, get_transformer_layer_cls, ignored_modules=None)
flow_grpo/fsdp_utils.py:39
↓ 6 callersFunctiongather_tensor
(tensor, world_size)
scripts/train_qwenimage.py:44
↓ 5 callersFunction_load_lines
Load lines from a file. First tries to load from `path` directly, and if that doesn't exist, searches the `flow_grpo/assets` directory for a
flow_grpo/prompts.py:13
↓ 5 callersFunctioncalculate_shift
( image_seq_len, base_seq_len: int = 256, max_seq_len: int = 4096, base_shift: float = 0.5,
flow_grpo/diffusers_patch/flux_pipeline_with_logprob.py:9
↓ 5 callersFunctionfrom_file
(path, low=None, high=None)
flow_grpo/prompts.py:27
↓ 5 callersFunctiongather_tensor
(tensor, world_size)
scripts/train_qwenimage_pr.py:46
↓ 5 callersFunctionset_seed
(seed, device_specific=True)
scripts/train_qwenimage.py:52
↓ 4 callersFunctioncompute_text_embeddings
(prompt, text_encoders, tokenizers, max_sequence_length, device)
scripts/train_sd3_pr.py:123
↓ 4 callersFunctioncompute_text_embeddings
(prompt, text_encoders, tokenizers, max_sequence_length, device)
scripts/train_sd3.py:123
↓ 4 callersFunctionencode_prompt
( text_encoders, tokenizers, prompt: str, max_sequence_length, device=None, num_images
flow_grpo/diffusers_patch/train_dreambooth_lora_sd3.py:98
↓ 3 callersFunctionpipeline_with_logprob
( self, prompt: Union[str, List[str]] = None, negative_prompt: Union[str, List[str]] = None, t
flow_grpo/diffusers_patch/qwenimage_pipeline_with_logprob.py:10
↓ 3 callersFunctionpipeline_with_logprob
r""" Function invoked when calling the pipeline for generation. Args: prompt (`str` or `List[str]`, *optional*): The promp
flow_grpo/diffusers_patch/sd3_pipeline_with_logprob.py:13
↓ 3 callersFunctionpipeline_with_logprob
( self, prompt: Union[str, List[str]] = None, prompt_2: Optional[Union[str, List[str]]] = None,
flow_grpo/diffusers_patch/flux_pipeline_with_logprob.py:22
↓ 2 callersFunctioncompute_log_prob
(transformer, pipeline, sample, j, config, rank)
scripts/train_qwenimage_pr.py:206
↓ 2 callersFunctioncompute_log_prob
(transformer, pipeline, sample, j, embeds, pooled_embeds, config)
scripts/train_sd3_pr.py:189
↓ 2 callersFunctioncompute_log_prob
(transformer, pipeline, sample, j, embeds, pooled_embeds, config)
scripts/train_sd3.py:189
↓ 2 callersFunctioncompute_log_prob
(transformer, pipeline, sample, j, config)
scripts/train_flux_pr.py:182
↓ 2 callersFunctioncompute_log_prob
(transformer, pipeline, sample, j, config, rank)
scripts/train_qwenimage.py:204
↓ 2 callersFunctioncompute_log_prob
(transformer, pipeline, sample, j, config)
scripts/train_flux.py:181
↓ 2 callersFunctioncompute_text_embeddings
(prompt, text_encoders, tokenizers, max_sequence_length, device)
scripts/train_flux_pr.py:124
↓ 2 callersFunctioncompute_text_embeddings
(prompt, text_encoders, tokenizers, max_sequence_length, device)
scripts/train_flux.py:123
↓ 2 callersFunctionencode_prompt
( text_encoders, tokenizers, prompt: str, max_sequence_length, device=None, num_images
flow_grpo/diffusers_patch/train_dreambooth_lora_flux.py:107
↓ 2 callersFunctionget_transformer_layer_cls
()
scripts/train_qwenimage.py:328
↓ 2 callersFunctioninit_distributed
Initialize distributed training
flow_grpo/fsdp_utils.py:146
↓ 2 callersFunctionsave_fsdp_checkpoint
(save_dir, model, global_step, rank)
flow_grpo/fsdp_utils.py:90
↓ 2 callersMethodstep
(self, *args, **kwargs)
flow_grpo/fsdp_utils.py:134
↓ 1 callersMethod__init__
(self)
flow_grpo/aesthetic_scorer.py:14
↓ 1 callersFunction_encode_prompt_with_clip
( text_encoder, tokenizer, prompt: str, device=None, text_input_ids=None, num_images_p
flow_grpo/diffusers_patch/train_dreambooth_lora_flux.py:63
↓ 1 callersFunction_encode_prompt_with_clip
( text_encoder, tokenizer, prompt: str, device=None, text_input_ids=None, num_images_p
flow_grpo/diffusers_patch/train_dreambooth_lora_sd3.py:59
↓ 1 callersFunction_encode_prompt_with_t5
( text_encoder, tokenizer, max_sequence_length=512, prompt=None, num_images_per_prompt=1,
flow_grpo/diffusers_patch/train_dreambooth_lora_flux.py:19
↓ 1 callersFunction_encode_prompt_with_t5
( text_encoder, tokenizer, max_sequence_length, prompt=None, num_images_per_prompt=1,
flow_grpo/diffusers_patch/train_dreambooth_lora_sd3.py:19
↓ 1 callersFunction_extract_scores
(text_outputs)
flow_grpo/rewards.py:347
↓ 1 callersFunctioncalculate_shift
( image_seq_len, base_seq_len: int = 256, max_seq_len: int = 4096, base_shift: float = 0.5,
flow_grpo/diffusers_patch/flux_pipeline_with_logprob_perstep.py:9
↓ 1 callersFunctioncalculate_zero_std_ratio
Calculate the proportion of unique prompts whose reward standard deviation is zero. Args: prompts: List of prompts. gathered
scripts/train_qwenimage_pr.py:158
↓ 1 callersFunctioncalculate_zero_std_ratio
Calculates the proportion of rewards for each unique prompt whose standard deviation is zero. Parameters: prompts: List of prompts
scripts/train_sd3_pr.py:140
↓ 1 callersFunctioncalculate_zero_std_ratio
Calculates the proportion of rewards for each unique prompt whose standard deviation is zero. Parameters: prompts: List of prompts
scripts/train_sd3.py:140
↓ 1 callersFunctioncalculate_zero_std_ratio
Calculate the proportion of unique prompts whose reward standard deviation is zero. Args: prompts: List of prompts. gath
scripts/train_flux_pr.py:134
↓ 1 callersFunctioncalculate_zero_std_ratio
Calculate the proportion of unique prompts whose reward standard deviation is zero. Args: prompts: List of prompts. gath
scripts/train_qwenimage.py:156
↓ 1 callersFunctioncalculate_zero_std_ratio
Calculate the proportion of unique prompts whose reward standard deviation is zero. Args: prompts: List of prompts. gath
scripts/train_flux.py:133
↓ 1 callersFunctiondistribute_samples
(tasks, weights, total_samples)
dataset/merge_genevaltask.py:5
↓ 1 callersFunctioneval
(pipeline, test_dataloader, config, rank, local_rank, world_size, device, global_step, reward_fn, executor, au
scripts/train_qwenimage_pr.py:244
↓ 1 callersFunctioneval
(pipeline, test_dataloader, text_encoders, tokenizers, config, accelerator, global_step, reward_fn, executor,
scripts/train_sd3_pr.py:224
↓ 1 callersFunctioneval
(pipeline, test_dataloader, text_encoders, tokenizers, config, accelerator, global_step, reward_fn, executor,
scripts/train_sd3.py:224
↓ 1 callersFunctioneval
(pipeline, test_dataloader, text_encoders, tokenizers, config, accelerator, global_step, reward_fn, executor,
scripts/train_flux_pr.py:215
↓ 1 callersFunctioneval
(pipeline, test_dataloader, config, rank, local_rank, world_size, device, global_step, reward_fn, executor, au
scripts/train_qwenimage.py:244
↓ 1 callersFunctioneval
(pipeline, test_dataloader, text_encoders, tokenizers, config, accelerator, global_step, reward_fn, executor,
scripts/train_flux.py:214
↓ 1 callersFunctionevaluate_batch_image
(images, prompts)
flow_grpo/rewards.py:387
↓ 1 callersFunctionevaluate_image
(prompt, image)
flow_grpo/rewards.py:363
↓ 1 callersFunctionextract_scores
(output_text)
flow_grpo/qwenvl.py:16
↓ 1 callersFunctionfilter_train_metadata
(test_path, train_path, output_path)
dataset/geneval_filter_test.py:3
↓ 1 callersMethodget_current_decay
(self, optimization_step)
flow_grpo/ema.py:33
↓ 1 callersFunctionjpeg_incompressibility
()
flow_grpo/rewards.py:7
↓ 1 callersFunctionload_optimizer_states_to_gpu
Move optimizer states to GPU
flow_grpo/fsdp_utils.py:119
↓ 1 callersMethodload_state_dict
(self, state_dict: dict)
flow_grpo/ema.py:79
↓ 1 callersFunctionmain
()
flow_grpo/rewards.py:455
↓ 1 callersFunctionmain
()
flow_grpo/stat_tracking.py:44
↓ 1 callersFunctionmain
()
flow_grpo/imagereward_scorer.py:21
↓ 1 callersFunctionmain
()
flow_grpo/qwenvl.py:102
↓ 1 callersFunctionmain
()
flow_grpo/pickscore_scorer.py:53
↓ 1 callersFunctionmerge_datasets_with_weights
(tasks, weights, output_path, total_samples=50000)
dataset/merge_genevaltask.py:35
↓ 1 callersFunctionmulti_score
(device, score_dict)
flow_grpo/rewards.py:409
↓ 1 callersFunctionoffload_optimizer_states_to_cpu
Move optimizer states to CPU
flow_grpo/fsdp_utils.py:108
↓ 1 callersFunctionpil_image_to_base64
(image)
flow_grpo/rewards.py:340
↓ 1 callersFunctionpil_image_to_base64
(image)
flow_grpo/qwenvl.py:9
↓ 1 callersFunctionread_jsonl
(file_path)
dataset/merge_genevaltask.py:22
↓ 1 callersFunctionsample_data
(data, samples_needed)
dataset/merge_genevaltask.py:29
↓ 1 callersFunctionsave_ckpt
(save_dir, transformer, global_step, accelerator, ema, transformer_trainable_parameters, config)
scripts/train_sd3_pr.py:330
↓ 1 callersFunctionsave_ckpt
(save_dir, transformer, global_step, accelerator, ema, transformer_trainable_parameters, config)
scripts/train_sd3.py:331
↓ 1 callersFunctionsave_ckpt
(save_dir, transformer, global_step, accelerator, ema, transformer_trainable_parameters, config)
scripts/train_flux_pr.py:311
↓ 1 callersFunctionsave_ckpt
(save_dir, transformer, global_step, accelerator, ema, transformer_trainable_parameters, config)
scripts/train_flux.py:310
↓ 1 callersMethodset_epoch
(self, epoch)
scripts/train_qwenimage_pr.py:144
↓ 1 callersMethodset_epoch
(self, epoch)
scripts/train_sd3_pr.py:120
↓ 1 callersMethodset_epoch
(self, epoch)
scripts/train_sd3.py:120
↓ 1 callersMethodset_epoch
(self, epoch)
scripts/train_flux_pr.py:120
↓ 1 callersMethodset_epoch
(self, epoch)
scripts/train_qwenimage.py:142
↓ 1 callersMethodset_epoch
(self, epoch)
scripts/train_flux.py:119
↓ 1 callersFunctionset_seed
(seed, device_specific=True)
scripts/train_qwenimage_pr.py:54
↓ 1 callersMethodstate_dict
(self)
flow_grpo/ema.py:84
↓ 1 callersFunctionunwrap_model
(model, accelerator)
scripts/train_sd3_pr.py:325
↓ 1 callersFunctionunwrap_model
(model, accelerator)
scripts/train_sd3.py:326
↓ 1 callersFunctionunwrap_model
(model, accelerator)
scripts/train_flux_pr.py:306
↓ 1 callersFunctionunwrap_model
(model, accelerator)
scripts/train_flux.py:305
Method__call__
(self, images)
flow_grpo/aesthetic_scorer.py:46
Method__call__
Calculate OCR reward :param images: List of input images (PIL or numpy format) :param prompts: Corresponding target text list
flow_grpo/ocr.py:22
Method__call__
(self, prompt, images)
flow_grpo/imagereward_scorer.py:16
Method__call__
(self, prompt, images)
flow_grpo/qwenvl.py:61
Method__call__
(self, prompt, images)
flow_grpo/pickscore_scorer.py:17
Method__getattr__
(self, name)
flow_grpo/fsdp_utils.py:140
Method__getitem__
(self, idx)
scripts/train_qwenimage_pr.py:75
Method__getitem__
(self, idx)
scripts/train_qwenimage_pr.py:94
Method__getitem__
(self, idx)
scripts/train_sd3_pr.py:56
next →1–100 of 220, ranked by callers