Functions — github.com/Agent-RL/ReCall

hub / github.com/Agent-RL/ReCall / functions

Functions1,539 in github.com/Agent-RL/ReCall

⨍ Functions 1,539◇ Types & classes 237↳ Endpoints 28

↓ 3 callersMethodto_dict

(self)

src/verl/single_controller/base/worker.py:78

↓ 3 callersMethodtoken_level_scores

(self, prediction: str, ground_truths: list)

src/flashrag/evaluator/metrics.py:56

↓ 3 callersMethodupdate_sampling_params

(self, **kwargs)

src/verl/workers/rollout/vllm_rollout/vllm_rollout_spmd.py:159

↓ 3 callersFunctionvocab_parallel_log_probs_from_logits

TODO(zhangchi.usc1992): We may change the implementation later

src/verl/utils/megatron/tensor_parallel.py:147

↓ 2 callersMethod__init__

(self, config)

src/verl/workers/megatron_workers.py:490

↓ 2 callersMethod__init__

(self, config)

src/verl/workers/fsdp_workers.py:635

↓ 2 callersMethod__init__

(self, input_size, num_heads, num_key_value_heads,

src/verl/models/llama/megatron/layers/parallel_linear.py:23

↓ 2 callersMethod__merge_token_to_word

(self, tokens, token_probs, force_tokens, token_map, force_reserve_digit)

src/flashrag/refiner/llmlingua_compressor.py:2190

↓ 2 callersMethod__token_prob_to_word_prob

(self, token_probs, convert_mode="mean")

src/flashrag/refiner/llmlingua_compressor.py:2217

↓ 2 callersFunction_broadcast_tp_shard_tensor_qkv

broadcast tensor in tp shards across mp_group

src/verl/models/qwen2/megatron/checkpoint_utils/qwen2_loader_depracated.py:281

↓ 2 callersFunction_broadcast_tp_shard_tensor_qkv

broadcast tensor in tp shards across mp_group

src/verl/models/qwen2/megatron/checkpoint_utils/qwen2_saver.py:259

↓ 2 callersFunction_broadcast_tp_shard_tensor_qkv

broadcast tensor in tp shards across mp_group

src/verl/models/mcore/saver.py:273

↓ 2 callersFunction_broadcast_tp_shard_tensor_qkv

broadcast tensor in tp shards across mp_group

src/verl/models/mcore/loader.py:281

↓ 2 callersMethod_build_model_optimizer

(self, model_path, megatron_config: ModelParalle

src/verl/workers/megatron_workers.py:137

↓ 2 callersMethod_build_model_optimizer

(self, model_path, fsdp_config,

src/verl/workers/fsdp_workers.py:141

↓ 2 callersMethod_compute_loss_and_backward

Compute loss with optional sequence parallelism and remove padding features

src/verl/trainer/fsdp_sft_trainer.py:290

↓ 2 callersFunction_compute_response_info

(batch: DataProto)

src/verl/trainer/ppo/metric_utils.py:32

↓ 2 callersMethod_download

(self, use_origin_parquet=False)

src/verl/utils/dataset/rl_dataset.py:119

↓ 2 callersFunction_fetch_tp_shard_tensor_qkv

fetch tensor in tp shards across mp_group

src/verl/models/qwen2/megatron/checkpoint_utils/qwen2_loader.py:164

↓ 2 callersMethod_forward_micro_batch

(self, micro_batch)

src/verl/workers/critic/dp_critic.py:50

↓ 2 callersMethod_forward_micro_batch

Returns: entropy: # (bs, response_len) log_probs: # (bs, response_len)

src/verl/workers/actor/dp_actor.py:61

↓ 2 callersFunction_get_ngrams

Extracts all n-grams upto a given maximum order from an input segment. Args: segment: text segment from which n-grams will be extracted.

src/flashrag/evaluator/_bleu.py:129

↓ 2 callersFunction_is_frac

(expr: str)

src/verl/utils/reward_score/prime_math/__init__.py:107

↓ 2 callersMethod_load_file_config

(self, config_file_path: str)

src/flashrag/config/config.py:48

↓ 2 callersFunction_normalize

Normalize answer expressions.

src/verl/utils/reward_score/prime_math/__init__.py:147

↓ 2 callersMethod_offload_params_to_cpu

(self, pp_rank, to_empty=False)

src/verl/workers/sharding_manager/megatron_vllm.py:123

↓ 2 callersMethod_pack_re_call_input

(self, prompt_template, func_schemas, user_input)

src/verl/utils/dataset/rl_dataset.py:125

↓ 2 callersFunction_pad_tensor

(x: Tensor, dim: int, padding_size: int)

src/verl/utils/ulysses.py:103

↓ 2 callersMethod_pad_to_length

(self, input_ids, attention_mask)

src/verl/utils/dataset/rm_dataset.py:99

↓ 2 callersFunction_parse

(s)

src/verl/utils/reward_score/prime_math/grader.py:312

↓ 2 callersMethod_post_process_params

For each param, if it is a tp-splited param, we all-gather from train tp group (vllm 0.8.2) or micro-dp group (vllm <= 0.6.3)

src/verl/workers/sharding_manager/megatron_vllm.py:371

↓ 2 callersFunction_pre_process_inputs

(pad_token_id, prompt_token_ids: torch.Tensor)

src/verl/workers/rollout/vllm_rollout/vllm_rollout_spmd.py:51

↓ 2 callersMethod_read_files_and_tokenize

(self)

src/verl/utils/dataset/rl_dataset.py:131

↓ 2 callersMethod_set_cos_sin_cache

(self, seq_len, device, dtype)

src/verl/models/qwen2/megatron/layers/parallel_attention.py:51

↓ 2 callersFunction_strip_properly_formatted_commas

(expr: str)

src/verl/utils/reward_score/prime_math/__init__.py:136

↓ 2 callersFunction_unpad_tensor

(x: Tensor, dim: int, padding_size: int)

src/verl/utils/ulysses.py:110

↓ 2 callersFunction_unwrap_ray_remote

(cls)

src/verl/single_controller/ray/base.py:454

↓ 2 callersMethod_validate

(self)

src/verl/trainer/ppo/ray_trainer.py:505

↓ 2 callersFunctionall_to_all_tensor

( local_input: Tensor, scatter_dim: int, gather_dim: int, group: Optional[dist.ProcessGroup] =

src/verl/utils/ulysses.py:132

↓ 2 callersFunctionallgather_dict_tensors

TODO: optimize this. - We can use async ops - We can use only one allgather Args: tensors: size: group:

src/verl/utils/torch_functional.py:191

↓ 2 callersMethodbatch_run

(self, dataset, batch_size=None)

src/flashrag/judger/judger.py:33

↓ 2 callersFunctionbootstrap_metric

(data: list[Any], subset_size: int, reduce_fns: list[Callable[[np.nd

src/verl/trainer/ppo/metric_utils.py:173

↓ 2 callersFunctioncompute_bleu

Computes BLEU score of translated segments against one or more references. Args: reference_corpus: list of lists of references for each tra

src/flashrag/evaluator/_bleu.py:149

↓ 2 callersFunctioncompute_policy_loss

Adapted from https://github.com/huggingface/trl/blob/main/trl/trainer/ppo_trainer.py#L1122 Args: old_log_prob: `(torch.Tensor)`

src/verl/trainer/ppo/core_algos.py:337

↓ 2 callersFunctioncompute_response_mask

(data: DataProto)

src/verl/trainer/ppo/ray_trainer.py:170

↓ 2 callersMethodconvert_ids_to_tokens

Converts a single index or a sequence of indices in a token or a sequence of tokens, using the vocabulary and added tokens.

src/verl/workers/rollout/tokenizer.py:116

↓ 2 callersFunctioncopy_local_path_from_hdfs

Deprecated. Please use copy_to_local instead.

src/verl/utils/fs.py:72

↓ 2 callersFunctioncreate_and_sync_state

(param_name, state, is_param)

src/verl/utils/fsdp_utils.py:285

↓ 2 callersMethodcritic_preds

Evaluate predictions using different retrieval docs

src/flashrag/pipeline/active_pipeline.py:329

↓ 2 callersMethodevaluate

The evaluation process after finishing overall generation

src/flashrag/pipeline/mm_pipeline.py:23

↓ 2 callersMethodexecute_all_async

(self, method_name: str, *args, **kwargs)

src/verl/single_controller/ray/base.py:375

↓ 2 callersMethodforward

(self, input_ids=None, attention_mask=None, **kwargs)

src/flashrag/generator/fid.py:32

↓ 2 callersMethodforward

Args: input_ids: input ids. shape (batch_size, seq_length) attention_mask: attention_mask. shape (batch_size, seq_le

src/verl/models/qwen2/megatron/modeling_qwen2_megatron.py:121

↓ 2 callersMethodforward

Args: input_ids: input ids. shape (batch_size, seq_length) attention_mask: attention_mask. shape (batch_size, seq_le

src/verl/models/llama/megatron/modeling_llama_megatron.py:121

↓ 2 callersMethodforward_backward_batch

(self, data: DataProto, forward_only=False)

src/verl/workers/critic/megatron_critic.py:116

↓ 2 callersMethodforward_backward_batch

We assume: - The model takes input: (input_ids, attention_mask, position_ids). No rmpad for the input - The communication sha

src/verl/workers/actor/megatron_actor.py:240

↓ 2 callersMethodfrom_single_dict

(cls, data: Dict[str, Union[torch.Tensor, np.ndarray]], meta_info=None)

src/verl/protocol.py:304

↓ 2 callersMethodget_checkpoint_name

Determine the directory name for this rank's checkpoint.

src/verl/utils/checkpoint/megatron_checkpoint_manager.py:125

↓ 2 callersFunctionget_common_default_kwargs_for_parallel_linear

()

src/verl/utils/megatron/tensor_parallel.py:47

↓ 2 callersMethodget_condition_ppl

( self, text: str, question: str, condition_in_question: str = "none",

src/flashrag/refiner/llmlingua_compressor.py:1070

↓ 2 callersFunctionget_constant_schedule_with_warmup

( optimizer: Optimizer, num_warmup_steps: int, last_epoch: int = -1, )

src/verl/utils/torch_functional.py:458

↓ 2 callersMethodget_estimate_threshold_base_distribution

(self, ppl, ratio: float, condition_flag: bool = False)

src/flashrag/refiner/llmlingua_compressor.py:1533

↓ 2 callersFunctionget_generation_config

( model: str, trust_remote_code: bool = False, )

src/verl/utils/model.py:58

↓ 2 callersFunctionget_hf_model_checkpoint_path

(checkpoint_path)

src/verl/utils/megatron_utils.py:277

↓ 2 callersFunctionget_huggingface_actor_config

(model_name: str, override_config_kwargs=None, trust_remote_code=False)

src/verl/utils/model.py:47

↓ 2 callersFunctionget_judger

(config)

src/flashrag/utils/utils.py:114

↓ 2 callersFunctionget_megatron_optimizer

( model, config: OptimizerConfig, no_weight_decay_cond=None, scale_lr_cond=Non

src/verl/utils/megatron/optimizer.py:27

↓ 2 callersFunctionget_model_checkpoint_path

(checkpoint_path)

src/verl/utils/megatron_utils.py:272

↓ 2 callersFunctionget_model_config

(model)

src/verl/utils/megatron_utils.py:38

↓ 2 callersMethodget_next_sentence

(self, output, scores)

src/flashrag/pipeline/active_pipeline.py:828

↓ 2 callersFunctionget_optimizer_checkpoint_path

(checkpoint_path, use_distributed_optimizer=True)

src/verl/utils/megatron_utils.py:282

↓ 2 callersFunctionget_parallel_gptmodel_from_config

(tfconfig, hf_config, pre_process=

src/verl/utils/model.py:426

↓ 2 callersFunctionget_pure_token

(token, model_name)

src/flashrag/refiner/llmlingua_compressor.py:100

↓ 2 callersMethodget_rank_results

( self, context: list, question: str, rank_method: str, condition_in_q

src/flashrag/refiner/llmlingua_compressor.py:1801

↓ 2 callersFunctionget_reranker

(config)

src/flashrag/utils/utils.py:103

↓ 2 callersFunctionget_rng_states_checkpoint_path

(checkpoint_path, only_rank0_save=True)

src/verl/utils/megatron_utils.py:294

↓ 2 callersFunctionget_seqlen_balanced_partitions

get order of seq lengths to make partitions balanced, this is used in balacing sum of seqlength across dp ranks and microbatches Paramete

src/verl/utils/seqlen_balancing.py:152

↓ 2 callersMethodget_string_with_varying_examplars

Select the maximum number of examplars that can be placed in the prompt

src/flashrag/prompt/base_prompt.py:142

↓ 2 callersFunctionget_token_params

Extract max_tokens and max_new_tokens from a parameter dictionary.

src/flashrag/generator/utils.py:18

↓ 2 callersFunctionget_trajectory_tracker

()

src/verl/utils/debug/trajectory_tracker.py:79

↓ 2 callersFunctionget_version

(pkg)

src/verl/workers/rollout/vllm_rollout/__init__.py:23

↓ 2 callersFunctionget_weight_buffer_meta_from_module

Return a dictionary containing name to a shape and dtype.

src/verl/utils/memory_buffer.py:61

↓ 2 callersFunctiongptmodel_forward

(model, input_ids, attention_mask, position_ids

src/verl/models/mcore/gpt_model.py:24

↓ 2 callersFunctiongrade_answer

The answer will be considered correct if: (a) it normalizes to the same string as the ground truth answer OR (b) sympy can simplify t

src/verl/utils/reward_score/prime_math/__init__.py:269

↓ 2 callersFunctioninit_megatron_optim_config

(optim_config: Dict)

src/verl/utils/megatron_utils.py:203

↓ 2 callersFunctionis_tensor_parallel_param

(param)

src/verl/utils/megatron/tensor_parallel.py:86

↓ 2 callersFunctionis_vllm_available

()

src/verl/utils/import_utils.py:34

↓ 2 callersMethodjudge

(self, dataset)

src/flashrag/judger/judger.py:87

↓ 2 callersFunctionjudge_image

(x)

src/flashrag/retriever/utils.py:167

↓ 2 callersMethodjudge_retrieve

Calculate whether a retrieve is required based on the output probability of the special token in the model

src/flashrag/pipeline/active_pipeline.py:285

↓ 2 callersFunctionkl_penalty

Compute KL divergence given logprob and ref_logprob. Copied from https://github.com/huggingface/trl/blob/main/trl/trainer/ppo_trainer.py#L1104

src/verl/trainer/ppo/core_algos.py:450

↓ 2 callersFunctionlast_boxed_only_string

(string)

src/verl/utils/reward_score/re_call.py:19

↓ 2 callersFunctionlast_boxed_only_string

Extract the last LaTeX boxed expression from a string. Args: string: Input string containing LaTeX code Returns:

src/verl/utils/reward_score/math_dapo.py:21

↓ 2 callersFunctionlist_of_dict_to_dict_of_list

(list_of_dict: list[dict])

src/verl/protocol.py:101

↓ 2 callersFunctionload_extern_type

Load a external data type based on the file path and type name

src/verl/utils/import_utils.py:61

↓ 2 callersFunctionload_fsdp_optimizer

(optimizer, device_id)

src/verl/utils/fsdp_utils.py:164

↓ 2 callersFunctionload_image_from_source

(image_path)

src/flashrag/generator/utils.py:99

↓ 2 callersMethodlocal_mkdir

(path)

src/verl/utils/checkpoint/checkpoint_manager.py:86

↓ 2 callersMethodlog

(self, data, step)

src/verl/utils/logger/aggregate_logger.py:40

↓ 2 callersMethodmake_iterator

r"""Make an iterator from the DataProto. This is built upon that TensorDict can be used as a normal Pytorch dataset. See https://pytorch.org/t

src/verl/protocol.py:561

← previous next →201–300 of 1,539, ranked by callers