Functions — github.com/Kwai-Klear/AR-GRPO

CodeHub MCPcopy Create free account

hub / github.com/Kwai-Klear/AR-GRPO / functions

Functions650 in github.com/Kwai-Klear/AR-GRPO

⨍ Functions 650◇ Types & classes 121↳ Endpoints 2

Method_reward_func

(self, cont, sol, **kwargs)

img_gen_grpo_rewards.py:565

Method_reward_func

(self, cont, sol, **kwargs)

img_gen_grpo_rewards.py:583

Method_reward_func

(self, cont, sol, **kwargs)

img_gen_grpo_rewards.py:639

Method_reward_func

(self, cont, sol, **kwargs)

img_gen_grpo_rewards.py:644

Method_reward_func

(self, cont, sol, **kwargs)

img_gen_grpo_rewards.py:649

Function_sample_with_triton_kernel

( probs: torch.Tensor, logprobs: torch.Tensor, sampling_metadata: SamplingMetadata, sampling_t

autoregressive/serve/sampler.py:520

Method_save_checkpoint

(self, model, trial)

modified_grpo_trainer.py:1508

Method_save_checkpoint

(self, *args, **kwargs)

img_gen_grpo_trainer.py:32

Method_set_signature_columns_if_needed

modified_grpo_trainer.py:733

Method_should_modify_greedy_probs_inplace

Whether or not the sampler should modify the probability distribution of greedily-sampled tokens such that multinomial sampling would sample

autoregressive/serve/sampler.py:128

Method_verify_args

autoregressive/serve/llm_engine.py:318

Methodabort_request

Aborts a request(s) with the given ID. Args: request_id: The ID(s) of the request to abort. Details: - Refer

autoregressive/serve/llm_engine.py:439

(self, lora_request: LoRARequest)

autoregressive/serve/model_runner.py:966

(self, lora_request: LoRARequest)

autoregressive/serve/gpu_executor.py:170

(self, lora_request: LoRARequest)

autoregressive/serve/llm_engine.py:661

(self, lora_request: LoRARequest)

autoregressive/serve/worker.py:257

Functionaesthetic_score

(images, model=None)

reward_utils/aesthscore.py:66

Functionapply_rotary_emb

(x: torch.Tensor, freqs_cis: torch.Tensor)

autoregressive/serve/gpt_model.py:360

Functioncalculate_one_clip_score

(pred, target, pred_type, target_type, model=None)

reward_utils/clip_score.py:66

Methodcheck_health

autoregressive/serve/gpu_executor.py:181

Methodcheck_health

autoregressive/serve/llm_engine.py:670

Functionchild_exit

(server, worker)

benchmark/reward-server/gunicorn.conf.py:22

Functioncompute_deqascore

benchmark/reward-server/reward_server/deqa.py:15

Functioncompute_geneval

(images, metadatas, only_strict=False)

benchmark/reward-server/reward_server/gen_eval.py:320

Methodcompute_liger_loss

(self, unwrapped_model, inputs)

modified_grpo_trainer.py:1328

Functioncreate_deepspeed_config

utils/deepspeed.py:1

Functiondeqa_score

(images, sess=None, only_strict=False, batch_size = 64, url = "http://127.0.0.1:18086")

reward_utils/deqa_client.py:18

Methoddetermine_num_available_blocks

Determine the number of available KV blocks by invoking the underlying worker.

autoregressive/serve/gpu_executor.py:136

Methoddo_log_stats

Forced log when no requests active.

autoregressive/serve/llm_engine.py:583

autoregressive/serve/model_runner.py:81

Methodencode_img

autoregressive/modeling.py:60

Methodentropy_from_logits

Calculate entropy from logits.

modified_grpo_trainer.py:1410

Evaluate if new feature vectors are at the manifold.

fid_evaluation.py:287

Methodexecute_model

( self, seq_group_metadata_list: List[SequenceGroupMetadata], blocks_to_swap_in: Dict[

autoregressive/serve/gpu_executor.py:153

Methodexecute_model_async

( self, seq_group_metadata_list: List[SequenceGroupMetadata], blocks_to_swap_in: Dict[

autoregressive/serve/gpu_executor.py:189

Methodextra_repr

utils/drop_path.py:35

Methodfilter_text

lazy_dataset.py:146

Inputs the output of the encoder network z and maps it to a discrete one-hot vector that is the index of the closest embedding vector

tokenizer_lg/vqgan/quantize.py:34

(self, z, temp=None, rescale_logits=False, return_logits=False)

tokenizer_lg/vqgan/quantize.py:168

tokenizer_lg/vqgan/layer.py:28

tokenizer_lg/vqgan/layer.py:47

(self, x, temb)

tokenizer_lg/vqgan/layer.py:96

tokenizer_lg/vqgan/layer.py:147

tokenizer_lg/vqgan/layer.py:239

tokenizer_lg/vqgan/layer.py:339

tokenizer_lg/vqgan/model.py:85

tokenizer_lg/tokenizer_image/vq_model.py:57

tokenizer_lg/tokenizer_image/vq_model.py:105

tokenizer_lg/tokenizer_image/vq_model.py:173

tokenizer_lg/tokenizer_image/vq_model.py:215

tokenizer_lg/tokenizer_image/vq_model.py:299

tokenizer_lg/tokenizer_image/vq_model.py:327

tokenizer_lg/tokenizer_image/vq_model.py:374

tokenizer_lg/tokenizer_image/vq_model.py:389

Standard forward.

tokenizer_lg/tokenizer_image/discriminator.py:74

(self, input, reverse=False)

tokenizer_lg/tokenizer_image/discriminator.py:112

tokenizer_lg/tokenizer_image/discriminator.py:203

tokenizer_lg/tokenizer_image/discriminator.py:229

tokenizer_lg/tokenizer_image/discriminator.py:244

Standard forward.

tokenizer_lg/tokenizer_image/discriminator_patchgan.py:65

(self, input, reverse=False)

tokenizer_lg/tokenizer_image/discriminator_patchgan.py:103

(self, codebook_loss, inputs, reconstructions, optimizer_idx, global_step, last_layer=None, l

tokenizer_lg/tokenizer_image/vq_loss.py:117

tokenizer_lg/tokenizer_image/discriminator_stylegan.py:48

tokenizer_lg/tokenizer_image/discriminator_stylegan.py:74

tokenizer_lg/tokenizer_image/discriminator_stylegan.py:90

(self, input, target)

tokenizer_lg/tokenizer_image/lpips.py:83

tokenizer_lg/tokenizer_image/lpips.py:105

tokenizer_lg/tokenizer_image/lpips.py:142

utils/drop_path.py:32

reward_utils/aesthscore.py:32

(self,input_ids=None,attention_mask=None,logits_to_keep=None,eval_mode=True,**kwargs)

autoregressive/modeling.py:279

autoregressive/serve/gpt_model.py:73

autoregressive/serve/gpt_model.py:119

( self, x: torch.Tensor, positions: torch.Tensor, kv_cache: torch.Tensor,

autoregressive/serve/gpt_model.py:154

( self, x: torch.Tensor, positions: torch.Tensor, kv_cache: torch.Tensor,

autoregressive/serve/gpt_model.py:191

(self, x: torch.Tensor, positions: torch.Tensor, kv_cache: torch.Tensor, attn_metadata: AttentionMetadata)

autoregressive/serve/gpt_model.py:238

( self, input_ids: torch.Tensor=None, positions: torch.Tensor=None, kv_caches

autoregressive/serve/gpt_model.py:273

( self, logits: torch.Tensor, sampling_metadata: SamplingMetadata, )

autoregressive/serve/sampler.py:46

(self, labels, train, force_drop_ids=None)

autoregressive/models/gpt.py:78

(self, caption, train, force_drop_ids=None)

autoregressive/models/gpt.py:110

autoregressive/models/gpt.py:127

autoregressive/models/gpt.py:146

autoregressive/models/gpt.py:166

( self, x: torch.Tensor, freqs_cis: torch.Tensor = None, input_pos: Optional[torch.Tensor] =

autoregressive/models/gpt.py:207

( self, x: torch.Tensor, freqs_cis: torch.Tensor, start_pos: int, mask: Optional[torch.Tensor] = None)

autoregressive/models/gpt.py:253

( self, idx: torch.Tensor, cond_idx: torch.Tensor, # cond_idx_or_embed inpu

autoregressive/models/gpt.py:336

(self, input_ids: Optional[torch.Tensor] = None, attention_mask=None, generation_config: Optional[GenerationCo

autoregressive/modeling.py:240

Methodget_codebook_entry

(self, indices, shape, channel_first=True)

tokenizer_lg/vqgan/quantize.py:211

Methodget_tokenizer

autoregressive/serve/llm_engine.py:298

Methodget_tokenizer_for_seq

(self, sequence: Sequence)

autoregressive/serve/llm_engine.py:301

Methodget_train_dataloader

modified_grpo_trainer.py:751

Functionhinge_d_loss

(logits_real, logits_fake)

tokenizer_lg/tokenizer_image/vq_loss.py:14

Functionhinge_gen_loss

tokenizer_lg/tokenizer_image/vq_loss.py:35

Functionidentity

Do we really need docs for this?

modified_grpo_trainer.py:286

Functionimagereward_score

(images, prompts, model=None)

reward_utils/imagereward.py:29

Functioninference

benchmark/reward-server/app_deqa.py:22

Functioninference

benchmark/reward-server/app_geneval.py:22

Functioninit_model

reward_utils/maniqa.py:72

Methodinitialize_cache

Initialize the KV cache by invoking the underlying worker.

autoregressive/serve/gpu_executor.py:142

Methodlast_layer

tokenizer_lg/tokenizer_image/vq_model.py:170

← previous next →501–600 of 650, ranked by callers