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
↓ 197 callersFunctionprint
(*args, **kwargs)
utils/distributed.py:13
↓ 41 callersMethodload
(cls, path: str, arr_name: str)
fid_evaluation.py:503
↓ 26 callersMethodfrom_pretrained
(cls, name="vgg_lpips")
tokenizer_lg/tokenizer_image/lpips.py:75
↓ 25 callersMethodupdate
(self, input_pos, k_val, v_val)
autoregressive/models/gpt.py:177
↓ 15 callersMethod__init__
(self,**kwargs)
img_gen_grpo_rewards.py:561
↓ 14 callersMethodencode
(self, x)
tokenizer_lg/vqgan/model.py:69
↓ 11 callersMethodlog
(self, logs: dict[str, float], start_time: Optional[float] = None)
modified_grpo_trainer.py:1468
↓ 9 callersFunctionto_pil_image
(x)
img_gen_grpo_rewards.py:10
↓ 8 callersMethod__init__
(self, in_features, hidden_features, out_features)
autoregressive/models/gpt.py:119
↓ 8 callersFunctionbuild_dataset
(data_path,script_args)
lazy_dataset.py:184
↓ 8 callersFunctioncreate_logger
Create a logger that writes to a log file and stdout.
utils/logger.py:4
↓ 8 callersMethoddecode
(self, quant)
tokenizer_lg/vqgan/model.py:75
↓ 8 callersMethoddecode_code
(self, code_b, shape, channel_first=True)
tokenizer_lg/vqgan/model.py:80
↓ 8 callersMethodgenerate
Generates the completions for the input prompts. NOTE: This class automatically batches the given prompts, considering the memory con
autoregressive/serve/llm.py:138
↓ 8 callersFunctionwith_article
(name: str)
benchmark/geneval/prompts/create_prompts.py:19
↓ 7 callersMethod__init__
(self, config: ModelArgs)
tokenizer_lg/tokenizer_image/vq_model.py:29
↓ 7 callersFunctiondiscrete_reward_map
(value,value_list,reward_list)
img_gen_grpo_rewards.py:14
↓ 6 callersFunctionNormalize
(in_channels)
tokenizer_lg/vqgan/layer.py:13
↓ 6 callersMethodempty
(cls)
autoregressive/serve/model_runner.py:56
↓ 6 callersFunctioninit_distributed_mode
(args)
utils/distributed.py:20
↓ 6 callersMethodstep
Performs one decoding iteration and returns newly generated results. .. figure:: https://i.imgur.com/sv2HssD.png :alt: Overview o
autoregressive/serve/llm_engine.py:511
↓ 5 callersFunctionNormalize
(in_channels, norm_type='group')
tokenizer_lg/tokenizer_image/vq_model.py:359
↓ 5 callersMethod__init__
(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks, attn_resolutions, dropout=0.0, resam
tokenizer_lg/vqgan/layer.py:176
↓ 5 callersMethod__init__
(self, config: ModelArgs)
autoregressive/serve/gpt_model.py:128
↓ 5 callersMethodbatch_decode
(self, input_ids,**kwargs)
autoregressive/modeling.py:73
↓ 5 callersMethodfreeze_model
(self,excluded_list=None)
autoregressive/modeling.py:224
↓ 5 callersFunctionleaky_relu
(p=0.2)
tokenizer_lg/tokenizer_image/discriminator.py:250
↓ 5 callersFunctionleaky_relu
(p=0.2)
tokenizer_lg/tokenizer_image/discriminator_stylegan.py:96
↓ 5 callersFunctionnonlinearity
(x)
tokenizer_lg/vqgan/layer.py:8
↓ 5 callersMethodsample
( self, logits: torch.Tensor, sampling_metadata: SamplingMetadata, )
autoregressive/serve/gpt_model.py:302
↓ 4 callersMethod__init__
(self)
tokenizer_lg/tokenizer_image/discriminator.py:239
↓ 4 callersMethod_get_per_token_logps
(self, model, input_ids, attention_mask, logits_to_keep, batch_size=None)
modified_grpo_trainer.py:862
↓ 4 callersFunctionapply_rotary_emb_bs
(x: torch.Tensor, freqs_cis: torch.Tensor)
autoregressive/serve/gpt_model.py:373
↓ 4 callersFunctioncenter_crop_arr
Center cropping implementation from ADM. https://github.com/openai/guided-diffusion/blob/8fb3ad9197f16bbc40620447b2742e13458d2831/guided_diff
utils/data.py:4
↓ 4 callersFunctioncreate_model
(device=None)
reward_utils/aesthscore.py:60
↓ 4 callersFunctioncreate_session
()
reward_utils/deqa_client.py:10
↓ 4 callersFunctiongenerate
(model, cond, max_new_tokens, emb_masks=None, cond_ids=None, cfg_scale=1.0, cfg_interval=-1, return_probs=Fals
autoregressive/models/generate.py:129
↓ 4 callersFunctioninit_pipeline
(model_or_path="Qwen/Qwen2.5-VL-3B-Instruct",device=None,**kwargs)
reward_utils/qwen_vl_reward.py:45
↓ 4 callersFunctionnonlinearity
(x)
tokenizer_lg/tokenizer_image/vq_model.py:354
↓ 4 callersFunctionupdate_ema
Step the EMA model towards the current model.
utils/ema.py:5
↓ 3 callersMethod__init__
(self, use_dropout=True)
tokenizer_lg/tokenizer_image/lpips.py:55
↓ 3 callersMethod_prepare_inputs
( self, generation_batch: dict[str, Union[torch.Tensor, Any]] )
modified_grpo_trainer.py:974
↓ 3 callersFunctioncenter_crop_arr
Center cropping implementation from ADM. https://github.com/openai/guided-diffusion/blob/8fb3ad9197f16bbc40620447b2742e13458d2831/guided_diff
tokenizer_lg/validation/val_ddp.py:55
↓ 3 callersFunctioncompute_statistics_of_path
(path, model, batch_size, dims, device, num_workers=1)
reward_utils/fid.py:217
↓ 3 callersMethodencode
(self, x)
tokenizer_lg/tokenizer_image/vq_model.py:41
↓ 3 callersFunctionencode_base64_from_img
(img)
reward_utils/qwen_vl_reward.py:51
↓ 3 callersMethodget_text_embeddings
(self, texts)
language/t5.py:49
↓ 3 callersFunctioninit_client
(openai_api_key = "EMPTY", url="localhost", port="20004", interf="v1")
reward_utils/qwen_vl_reward.py:37
↓ 3 callersFunctioninit_clip
(device=None)
reward_utils/clip_score.py:46
↓ 3 callersMethodinit_from_ckpt
(self, path, ignore_keys=list(), logging=True)
tokenizer_lg/vqgan/model.py:55
↓ 3 callersFunctioninit_model
(device,dims=2048)
reward_utils/fid.py:257
↓ 3 callersFunctioninitialize_model
(ckpt_path: str = None, device=None, hps_version: str = "v2.0")
reward_utils/hpsv2.py:13
↓ 3 callersMethodload_text_encoder
(self)
autoregressive/modeling.py:195
↓ 3 callersMethodtext_embedding
(self,input_ids,attention_mask)
language/t5.py:88
↓ 2 callersMethod__init__
(self)
tokenizer_lg/tokenizer_image/discriminator_stylegan.py:85
↓ 2 callersFunction_batch_pairwise_distances
Compute pairwise distances between two batches of feature vectors.
fid_evaluation.py:418
↓ 2 callersFunction_beam_search_sample
( selected_seq_groups: List[Tuple[List[int], SamplingParams]], is_prompts: List[bool], seq_data: D
autoregressive/serve/sampler.py:325
↓ 2 callersFunction_download_inception_model
()
fid_evaluation.py:577
↓ 2 callersMethod_generate_and_score_completions
( self, inputs: list[dict[str, Union[torch.Tensor, Any]]] )
modified_grpo_trainer.py:1056
↓ 2 callersFunction_get_bin_counts_and_mask
( tokens: torch.Tensor, vocab_size: int, num_seqs: int, )
autoregressive/serve/sampler.py:143
↓ 2 callersFunction_get_graph_batch_size
Returns the padded batch size given actual batch size. Batch sizes are 1, 2, 4, _BATCH_SIZE_ALIGNMENT, 2*_BATCH_SIZE_ALIGNMENT, 3*_BATCH_SIZE
autoregressive/serve/model_runner.py:1202
↓ 2 callersMethod_get_stats
Get Stats to be Logged to Prometheus.
autoregressive/serve/llm_engine.py:588
↓ 2 callersFunction_greedy_sample
( selected_seq_groups: List[Tuple[List[int], SamplingParams]], samples: torch.Tensor, )
autoregressive/serve/sampler.py:279
↓ 2 callersFunction_maybe_pynccl
()
autoregressive/serve/model_runner.py:1193
↓ 2 callersFunction_random_sample
( selected_seq_groups: List[Tuple[List[int], SamplingParams]], is_prompts: List[bool], random_samp
autoregressive/serve/sampler.py:298
↓ 2 callersMethod_sync_fsdp_params_to_vllm
Memory-efficient post-order traversal of FSDP modules to extract full parameters and sync with vLLM.
modified_grpo_trainer.py:882
↓ 2 callersFunctionadopt_weight
(weight, global_step, threshold=0, value=0.)
tokenizer_lg/tokenizer_image/vq_loss.py:43
↓ 2 callersFunctionapply_rotary_emb
(x: torch.Tensor, freqs_cis: torch.Tensor)
autoregressive/models/gpt.py:436
↓ 2 callersFunctionbuild_model
(script_args, training_args, model_args)
img_gen_grpo_train.py:114
↓ 2 callersFunctioncalculate_fid
(imgs, refs=None, imgs_stat=None, refs_stat=None, model=None)
reward_utils/fid.py:263
↓ 2 callersFunctioncalculate_frechet_distance
Numpy implementation of the Frechet Distance. The Frechet distance between two multivariate Gaussians X_1 ~ N(mu_1, C_1) and X_2 ~ N(mu_2, C_2
reward_utils/fid.py:131
↓ 2 callersMethodcapture
( self, input_ids: torch.Tensor, positions: torch.Tensor, kv_caches: List[torc
autoregressive/serve/model_runner.py:1112
↓ 2 callersFunctioncenter_crop_arr
Center cropping implementation from ADM. https://github.com/openai/guided-diffusion/blob/8fb3ad9197f16bbc40620447b2742e13458d2831/guided_diff
lazy_dataset.py:15
↓ 2 callersMethodclean_caption
(self, caption)
language/t5.py:112
↓ 2 callersFunctioncolor_classification
(image, bboxes, classname)
benchmark/reward-server/reward_server/gen_eval.py:88
↓ 2 callersMethodcompute_activations
Compute image features for downstream evals. :param batches: a iterator over NHWC numpy arrays in [0, 255]. :return: a tuple
fid_evaluation.py:146
↓ 2 callersFunctioncompute_statistics
We use image batch as this function could also used for metric. For GRPO specificly, we only have 1 sample for 1 completion, there is no batc
reward_utils/fid.py:235
↓ 2 callersFunctioncreat_optimizer
(model, weight_decay, learning_rate, betas, logger)
autoregressive/train/train_c2i.py:28
↓ 2 callersMethoddecode
(self, quant)
tokenizer_lg/tokenizer_image/vq_model.py:47
↓ 2 callersMethoddecode
(self, input_ids, **kwargs)
autoregressive/modeling.py:63
↓ 2 callersMethoddecode_code
(self, code_b, shape=None, channel_first=True)
tokenizer_lg/tokenizer_image/vq_model.py:52
↓ 2 callersMethodevaluate_pr
Evaluate precision and recall efficiently. :param features_1: [N1 x D] feature vectors for reference batch. :param radii_1:
fid_evaluation.py:329
↓ 2 callersFunctionfind_multiple
(n: int, k: int)
autoregressive/models/gpt.py:18
↓ 2 callersFunctionforward_modality
(model, data, flag)
reward_utils/clip_score.py:82
↓ 2 callersMethodfrechet_distance
Compute the Frechet distance between two sets of statistics.
fid_evaluation.py:76
↓ 2 callersFunctiongeneval_score
(images, metadatas, sess=None, only_strict=False, batch_size = 64, url = "http://127.0.0.1:18085")
reward_utils/geneval_client.py:18
↓ 2 callersFunctionget_ckpt_path
(name, root, check=False)
tokenizer_lg/tokenizer_image/lpips.py:42
↓ 2 callersMethodget_tokenizer
( self)
autoregressive/serve/llm.py:128
↓ 2 callersMethodinit_device
(self)
autoregressive/serve/worker.py:89
↓ 2 callersMethodinitialize
(self, input)
tokenizer_lg/tokenizer_image/discriminator.py:91
↓ 2 callersMethodinitialize
(self, input)
tokenizer_lg/tokenizer_image/discriminator_patchgan.py:82
↓ 2 callersMethodmanifold_radii
(self, features: np.ndarray)
fid_evaluation.py:252
↓ 2 callersFunctionmd5_hash
(path)
tokenizer_lg/tokenizer_image/lpips.py:36
↓ 2 callersFunctionnormalize_tensor
(x,eps=1e-10)
tokenizer_lg/tokenizer_image/lpips.py:158
↓ 2 callersMethodpairwise_distances
Evaluate pairwise distances between two batches of feature vectors.
fid_evaluation.py:397
↓ 2 callersFunctionprecompute_freqs_cis_2d
(grid_size: int, n_elem: int, base: int = 10000, cls_token_num=120)
autoregressive/models/gpt.py:420
↓ 2 callersMethodread_activations
(self, npz_path: str)
fid_evaluation.py:142
↓ 2 callersMethodread_statistics
( self, npz_path: str, activations: Tuple[np.ndarray, np.ndarray] )
fid_evaluation.py:168
↓ 2 callersFunctionrelative_position
Give position of A relative to B, factoring in object dimensions
benchmark/reward-server/reward_server/gen_eval.py:119
↓ 2 callersFunctionrequires_grad
Set requires_grad flag for all parameters in a model.
utils/ema.py:17
next →1–100 of 650, ranked by callers