MCPcopy Create free account

hub / github.com/Gen-Verse/OpenClaw-RL / functions

Functions10,561 in github.com/Gen-Verse/OpenClaw-RL

↓ 2 callersFunction_make_log_tag
( task_meta: Dict[str, Any], run_ctx: RunContext, lease_id: str | None = None, )
terminal-rl/generate.py:37
↓ 2 callersFunction_mark_aborted_samples
Give ABORTED samples a default reward and exclude them from training. The rollout pipeline skips reward_func for lists containing any ABORTED
gui-rl/generate_with_gui.py:706
↓ 2 callersFunction_mask_loss
Apply mask to the unreduced loss tensor.
Megatron-LM/megatron/post_training/loss_func.py:13
↓ 2 callersFunction_maybe_update_timeouts
(is_closing_ft=False)
Megatron-LM/megatron/training/ft_integration.py:242
↓ 2 callersFunction_merge_tp_partitions
( name: str, param_partitions: list[torch.Tensor], partition_dim: int, partition_stride: int,
slime/slime/backends/megatron_utils/update_weight/common.py:18
↓ 2 callersMethod_move_book_keeping_tensors
Move all the relevent booking tensors with src idxs to dst idxs
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:1547
↓ 2 callersFunction_multicellrange_containsby
(subset_candidate: MultiCellRange, superset_candidate: MultiCellRange)
gui-rl/desktop_env/evaluators/metrics/utils.py:708
↓ 2 callersMethod_new_bucket
Helper function that creates a new bucket. Also updates param->bucket mapping.
Megatron-LM/megatron/core/distributed/param_and_grad_buffer.py:884
↓ 2 callersFunction_normalize
Minimal normalization - only handle basic formatting: - Skip obvious file metadata (encoding, shebang) at the beginning - Nor
gui-rl/desktop_env/evaluators/metrics/general.py:605
↓ 2 callersFunction_normalize
Normalize answer expressions.
slime/slime/rollout/rm_hub/math_utils.py:250
↓ 2 callersFunction_normalize_messages_for_template
Make messages compatible with the chat template. - developer → system (templates only know 'system') - multimodal content lists → plain text
openclaw-rl/openclaw_api_server.py:47
↓ 2 callersFunction_nvtx_range_get_func_path
Get the path of a function. Assumes being called from nvtx_range_push/pop. Returns: str: Module path and function name joined by a dot
Megatron-LM/megatron/core/utils.py:2120
↓ 2 callersFunction_offload_rollout_data_to_cpu
Move per-sample GPU tensors in rollout_data back to CPU to free GPU memory. After compute_advantages_and_returns, various computed tensors (log_p
slime/slime/backends/megatron_utils/actor.py:137
↓ 2 callersFunction_open_zarr_array_verbose
(path: Path, mode: str, **open_kwargs)
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:249
↓ 2 callersFunction_overlap_count_per_token
``O[k, t] = | S^q_t ∩ S^p_{t,k} |`` via broadcasted equality. Memory: ``[K, R, K_q, K_p]`` boolean intermediate. With K_q=K_p=20 and R≤8192,
openclaw-combine/hint_opd_select_loss.py:100
↓ 2 callersFunction_p_assert
Alternate to ``assert`` when in the backward context to print the error message ``s`` since otherwise, it is swallowed.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:104
↓ 2 callersFunction_pad_if_needed
(numel_unpadded, divisor)
Megatron-LM/tests/unit_tests/distributed/test_param_and_grad_buffer.py:91
↓ 2 callersMethod_pad_tensor_for_quantization
Padding tensor shape to multiples of 16/32.
Megatron-LM/megatron/core/transformer/moe/experts.py:1084
↓ 2 callersFunction_param_generator
(cpu_optimizer)
Megatron-LM/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py:8
↓ 2 callersMethod_parse_experience_items
Parse '- EXPERIENCE ITEM: ...' lines from generated text.
openclaw-rl/oel/openclaw_oel_api_server.py:1218
↓ 2 callersFunction_pick_from_mapping
(data: dict[str, Any], key_names: tuple[str, ...] | None)
slime/slime/utils/eval_config.py:69
↓ 2 callersFunction_pop_first
(buffer: list[list[Sample]], num_samples: int)
gui-rl/gui_data_source.py:18
↓ 2 callersFunction_post
(client, url, payload, max_retries=60)
slime/slime/utils/http_utils.py:179
↓ 2 callersMethod_preprocess_data
Preprocess input data before input to language model. This function is adopted from https://github.com/huggingface/transformers/blob/
Megatron-LM/megatron/core/models/multimodal/llava_model.py:430
↓ 2 callersFunction_prm_eval_majority_vote
(scores: list[int | None])
openclaw-opd/openclaw_opd_api_server.py:214
↓ 2 callersMethod_process_embedding_token_parallel
Processes the input data for model parallelism support. When using sequence parallelism (SP) or context parallelism (CP), the sequence is sha
Megatron-LM/megatron/core/models/multimodal/llava_model.py:662
↓ 2 callersFunction_process_memory
Get memory used by current process. Returns (int): memory used by current process
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:635
↓ 2 callersFunction_profile_simple_loop
(iterator, args, name)
slime/slime/utils/profile_utils.py:48
↓ 2 callersMethod_query
(self, messages: list[dict[str, str]], **kwargs)
swe-rl/mini-swe-agent/src/minisweagent/models/litellm_model.py:53
↓ 2 callersMethod_query_judge_once
(self, messages: list[dict], vote_id: int)
openclaw-tinker/scorers.py:389
↓ 2 callersMethod_query_prm_eval_once
(self, prompt_text: str, vote_id: int)
openclaw-opd/openclaw_opd_api_server.py:447
↓ 2 callersMethod_reap_idle_locked
(self)
terminal-rl/remote/pool_server.py:135
↓ 2 callersMethod_reestablish_shared_weights
Reestablishes shared (tied) weights in a PyTorch module after parameter replacement. When iterating over `named_parameters()`, PyTor
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/megatron_fsdp.py:1158
↓ 2 callersMethod_reformulate_tool_call
Reformulate tool call instruction for tau-bench environment. The default tool template assumes one or more function calls, but for
slime/examples/tau-bench/trainable_agents.py:75
↓ 2 callersFunction_render_observation
Render the action_observation_template from swebench.yaml.
swe-rl/generate_with_swe_trajectory.py:452
↓ 2 callersMethod_replace_param_with_raw_if_needed
(self)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/megatron_fsdp.py:1144
↓ 2 callersFunction_replace_state_dict_keys_with_sharded_keys
Group ShardedBase objects by keys and return mappings required for recreating the original dict.
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:414
↓ 2 callersMethod_request
( self, method: str, worker_url: str, path: str, payload: dict[str, An
terminal-rl/router_server.py:94
↓ 2 callersMethod_reset_parameters
(self, old_params, new_params)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:2334
↓ 2 callersFunction_reshard_if_dtensor
Reshards the input tensor to match the sharding configuration of the reference tensor if the reference tensor is a DTensor. Otherwise, return
Megatron-LM/megatron/core/distributed/finalize_model_grads.py:60
↓ 2 callersFunction_resolve_ip
Attempt to get the local LAN IP for the specific family (IPv4/IPv6). Strategy: UDP Probe (Preferred) -> Hostname Resolution (Fallback
slime/slime/utils/http_utils.py:51
↓ 2 callersMethod_restore_common_per_param_step
(state_dict: Dict, step: Union[int, torch.Tensor])
Megatron-LM/megatron/core/optimizer/optimizer.py:340
↓ 2 callersMethod_restore_state
Internal method that restores the state that was saved in _save_state().
Megatron-LM/megatron/core/rerun_state_machine.py:908
↓ 2 callersFunction_round
Round `batch_size` down to nearest batch size divisible by `divisor`.
Megatron-LM/megatron/core/num_microbatches_calculator.py:261
↓ 2 callersMethod_run_full_checkpoint_test
(self, fp8)
Megatron-LM/tests/unit_tests/transformer/test_transformer_block.py:86
↓ 2 callersMethod_run_parameter_update_test
Helper method to test parameter updates with a given model class. Args: model_class: Model class to use for testing
Megatron-LM/tests/unit_tests/test_layer_wise_optimizer.py:356
↓ 2 callersMethod_run_selective_checkpoint_test
(self, fp8)
Megatron-LM/tests/unit_tests/transformer/test_transformer_block.py:117
↓ 2 callersMethod_run_step
(cls, env)
Megatron-LM/tests/unit_tests/inference/engines/test_dynamic_engine.py:431
↓ 2 callersMethod_run_test_helper
Test fp8_param with gpt_model.
Megatron-LM/tests/unit_tests/transformer/test_cuda_graphs.py:1160
↓ 2 callersFunction_s3_object_exists
Ascertain whether the object at the given S3 path exists in S3 Args: client (S3Client): The S3 client path (str): The S3 path
Megatron-LM/megatron/core/datasets/object_storage_utils.py:135
↓ 2 callersFunction_sanitize
(lst: list[float], name: str, session_id: str, turn_num: int)
openclaw-tinker/data_formatter.py:62
↓ 2 callersMethod_sanitize_data_iterators
( self, data_iterator: DataIteratorArgType )
Megatron-LM/megatron/core/rerun_state_machine.py:857
↓ 2 callersMethod_save_debug_rollout_data
(self, data, rollout_id, evaluation: bool)
slime/slime/ray/rollout.py:320
↓ 2 callersFunction_select_k_star_per_token
Compute ``k*(t) ∈ [0, K)`` per response token. * ``shortest`` : ``k* = 0`` for every token (the rollout module orders candidates sh
openclaw-combine/hint_opd_select_loss.py:113
↓ 2 callersMethod_set
(cls, schema, model, params)
Megatron-LM/tools/checkpoint/schema_base.py:76
↓ 2 callersFunction_set_capture_end
Set graph capture has ended.
Megatron-LM/megatron/core/transformer/cuda_graphs.py:89
↓ 2 callersFunction_set_capture_start
Set graph capture has started.
Megatron-LM/megatron/core/transformer/cuda_graphs.py:83
↓ 2 callersMethod_set_cos_sin_cache
(self, seq_len, offset, dtype, packed_seq_params=None)
Megatron-LM/megatron/core/models/common/embeddings/yarn_rotary_pos_embedding.py:187
↓ 2 callersMethod_setup
(self)
Megatron-LM/tools/retro/config_utils.py:277
↓ 2 callersFunction_shard_and_copy_
Copy weights from *ref_block* into a tensor-parallel *tgt_block*.
Megatron-LM/tests/unit_tests/pipeline_parallel/test_bridge_communicator.py:73
↓ 2 callersFunction_sharded_object_id
Unique id of the sharded object data. Should yield the same value for same data replicated on different ranks. Args: sharded_object
Megatron-LM/megatron/core/dist_checkpointing/utils.py:55
↓ 2 callersMethod_sharded_state_dict_grouped
prefix should be module_name to make keys identical to sequetial ones.
Megatron-LM/megatron/core/extensions/kitchen.py:705
↓ 2 callersMethod_sharded_state_dict_grouped
prefix should be module_name to make keys identical to sequetial ones.
Megatron-LM/megatron/core/extensions/transformer_engine.py:1499
↓ 2 callersFunction_should_create_array
(ten: ShardedTensor)
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:139
↓ 2 callersMethod_slice_conv_param
Slices a cp_size=1 conv1d parameter (either weight or bias) along the first dimension, returning the parts of the parameter needed fo
Megatron-LM/megatron/core/ssm/mamba_context_parallel.py:234
↓ 2 callersFunction_split_tensor_factory
Builds a factory that splits a given ShardedTensor into several independent chunks.
Megatron-LM/megatron/core/ssm/gated_delta_net.py:517
↓ 2 callersFunction_split_tensor_factory
Builds a factory that splits a given ShardedTensor into several independent chunks.
Megatron-LM/megatron/core/ssm/mamba_mixer.py:1146
↓ 2 callersMethod_start_emulator
(self)
gui-rl/desktop_env/desktop_env.py:198
↓ 2 callersFunction_start_router
Start a router for rollout or PRM engines.
slime/slime/ray/rollout.py:1131
↓ 2 callersFunction_status_from_payload
(payload: dict[str, Any], default: int)
terminal-rl/router_server.py:28
↓ 2 callersFunction_strip_chain_of_thought
(text: str)
slime/slime/rollout/rm_hub/gpqa.py:8
↓ 2 callersFunction_strip_properly_formatted_commas
(expr: str)
slime/slime/rollout/rm_hub/math_utils.py:239
↓ 2 callersMethod_submit_opd_sample
(self, turn_data, session_id, opd_result, reward=0.0)
openclaw-tinker/api_server.py:1033
↓ 2 callersMethod_submit_turn_sample
( self, turn_data: dict[str, Any], session_id: str, opd_result: dict[str, Any]
openclaw-fireworks/fireworks_server.py:585
↓ 2 callersMethod_submit_turn_sample
( self, turn_data: dict[str, Any], session_id: str, opd_result: dict[str, Any]
openclaw-combine/openclaw_combine_api_server.py:40
↓ 2 callersMethod_sync_hdo_param_groups_to_sub_optimizers
Sync HDO new param_groups attribute (e.g. lr, wd, etc.) to sub-optimizers.
Megatron-LM/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py:334
↓ 2 callersMethod_synchronize_steps
Synchronize the step of all optimizers. TE FusedAdam will not accumulate "step" for empty param groups, so we need to align t
Megatron-LM/megatron/core/optimizer/optimizer.py:1392
↓ 2 callersFunction_test_interruption_helper
Helper function for testing interruption scenarios.
swe-rl/mini-swe-agent/tests/agents/test_interactive.py:125
↓ 2 callersMethod_text_to_ids_extra_space
Converts text to tokens ids.
Megatron-LM/megatron/core/tokenizers/text/libraries/sentencepiece_tokenizer.py:204
↓ 2 callersFunction_throttle_by_lockfile
Ensure any two calls guarded by the same lock_file are spaced by min_interval across all local processes (and also across nodes if lock_file
gui-rl/desktop_env/providers/volcengine/manager.py:79
↓ 2 callersFunction_tinker_teacher_logprobs
Query Tinker teacher model for per-token logprobs on the student's response. Args: normalize_fn: Function to normalize messages (e.g., _n
openclaw-tinker/scorers.py:212
↓ 2 callersMethod_torch_sampling_func
Samples the logits to generate outputs Given the logits of the last token, this function samples it according to the parameters defin
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:218
↓ 2 callersFunction_truncate
(text: str, limit: int = 2000)
terminal-rl/agent/prm_agent.py:66
↓ 2 callersFunction_truncate
Truncate text and add ellipsis if too long.
slime/slime/utils/debug_utils/replay_reward_fn.py:12
↓ 2 callersMethod_truncate_experience
Truncate experience to max_tokens, keeping the most recent items.
openclaw-rl/oel/openclaw_oel_api_server.py:1229
↓ 2 callersFunction_update_budget
(budget, consumed: int)
slime/examples/geo3k_vlm_multi_turn/rollout.py:294
↓ 2 callersMethod_update_expert_bucket_weights_from_distributed
Gather EP → HF → broadcast. Clears buffer.
slime/slime/backends/megatron_utils/update_weight/update_weight_from_distributed.py:182
↓ 2 callersMethod_update_top_n_logprobs_dict
Function to update the top_n_logprobs at each step This function goes through the topn logprobs generated for each, and for whichever
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:957
↓ 2 callersMethod_update_total_costs
(self)
swe-rl/mini-swe-agent/src/minisweagent/run/extra/utils/batch_progress.py:117
↓ 2 callersMethod_validate_global_shapes
(self, metadata, sharded_tensors)
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:597
↓ 2 callersMethod_validate_params
(algo)
Megatron-LM/megatron/core/dist_checkpointing/tensor_aware_state_dict.py:61
↓ 2 callersFunction_vocab_parallel_topk_log_probs
Global top-K log-probs and (global vocab) indices from TP-sharded logits. Args: logits: ``[seq, vocab_local]`` (vocab_local = vocab // tp
slime/slime/backends/megatron_utils/loss.py:249
↓ 2 callersFunction_w_opd
()
openclaw-combine/hint_opd_loss.py:230
↓ 2 callersFunction_w_rl
()
openclaw-combine/hint_opd_loss.py:226
↓ 2 callersFunction_wait_for_instance_running
Wait for instance to reach Running state
gui-rl/desktop_env/providers/aliyun/manager.py:217
↓ 2 callersFunction_wait_server_healthy
(base_url, api_key, is_process_alive)
slime/slime/backends/sglang_utils/sglang_engine.py:82
↓ 2 callersFunction_wrap_ipv6
Wrap IPv6 address in [] if needed.
slime/slime/utils/http_utils.py:108
↓ 2 callersFunction_wrap_low_level_call
()
slime/slime/utils/reloadable_process_group.py:276
↓ 2 callersFunction_yarn_find_correction_dim
( num_rotations: float, dim: int, rotary_base: float = 10000, max_position_embeddings: int = 2048 )
Megatron-LM/megatron/core/models/common/embeddings/yarn_rotary_pos_embedding.py:219
↓ 2 callersFunction_yarn_get_concentration_factor
Get the concentration factor (factor multiplied to the sine and cosine components of the embedding). This factor is also known as attention f
Megatron-LM/megatron/core/models/common/embeddings/yarn_rotary_pos_embedding.py:260
← previousnext →1,601–1,700 of 10,561, ranked by callers