Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/Gen-Verse/OpenClaw-RL
/ functions
Functions
10,561 in github.com/Gen-Verse/OpenClaw-RL
⨍
Functions
10,561
◇
Types & classes
1,405
↳
Endpoints
144
↓ 1 callers
Function
offload_train
()
slime/train.py:38
↓ 1 callers
Method
on_dynamic_filter_drop
(self, reason: str | None)
slime/slime/rollout/filter_hub/base_types.py:28
↓ 1 callers
Method
on_eval
(self, metrics: dict[str, float])
slime/slime/utils/metric_utils.py:137
↓ 1 callers
Method
on_group_commit_backward
Called at the end of a layer group's backward pass. Ensures correct chunk is active and synchronizes reloads.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:441
↓ 1 callers
Method
on_group_commit_forward
Called at the end of a layer group's forward pass to trigger offloading.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:420
↓ 1 callers
Method
on_group_start_backward
Called at the start of a layer group's backward pass. Triggers reloading of tensors from CPU.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:469
↓ 1 callers
Method
on_group_start_forward
Called at the start of a layer group's forward pass. Increments group index and prepares for offloading.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:459
↓ 1 callers
Method
on_message_added
(self)
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:331
↓ 1 callers
Method
on_mount
Initialize the widget state.
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:153
↓ 1 callers
Function
on_save_checkpoint_start
Function to be called before save-checkpoint start to generate productive metrics to log after ckpt succeeds. Args: async_save (bool): ap
Megatron-LM/megatron/training/one_logger_utils.py:226
↓ 1 callers
Function
on_save_checkpoint_success
Function to be called after checkpointing succeeds and checkpoint is persisted for logging it as an artifact in W&B Args: checkpoint_path
Megatron-LM/megatron/training/wandb_utils.py:19
↓ 1 callers
Method
on_unmount
Clean up the log handler when the app shuts down.
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:343
↓ 1 callers
Function
onelogger_finalize_fn
()
Megatron-LM/megatron/training/checkpointing.py:708
↓ 1 callers
Function
otherLang
(lang)
slime/docs/_static/js/lang-toggle.js:9
↓ 1 callers
Function
override_sharded_param_cpu_function_closure
(p, cpu_function)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3736
↓ 1 callers
Function
override_sharded_param_methods_with_safety_checks
Override the methods of the parameters to prevent undefined behavior. Args: params (List[torch.Tensor]): The parameters to add hint o
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3711
↓ 1 callers
Function
override_sharded_param_to_function_closure
(p, to_function)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3722
↓ 1 callers
Function
overwrite_nccl_comm_cfgs
Overwrite the nccl_comm_cfgs for the given pg_name with the given key_value_pair.
Megatron-LM/megatron/core/parallel_state.py:535
↓ 1 callers
Function
pack_inference_logprobs
Pack inference logprobs into bins aligned with packed sequences. Args: inference_logprobs: List of inference logprobs tensors for each se
Megatron-LM/megatron/rl/rl_utils.py:207
↓ 1 callers
Function
pack_layer
(weight, group_size, sym=True)
slime/tools/convert_hf_to_int4_direct.py:121
↓ 1 callers
Function
pack_layer
(weight, group_size, sym=True)
slime/slime/backends/megatron_utils/megatron_to_hf/processors/quantizer_compressed_tensors.py:233
↓ 1 callers
Method
pack_selected_samples
Function to pack a list of ImageTaskSample into a single ImageTaskSamplePacked. NOTE: Energon dataloader calls this method internall
Megatron-LM/examples/multimodal/dataset_helpers.py:882
↓ 1 callers
Function
pack_sequences
Pack sequences into dense batches with cumulative sequence lengths. Args: tokens: List of token sequences loss_masks: List o
slime/slime/backends/fsdp_utils/data_packing.py:10
↓ 1 callers
Function
pack_to_int32
( value, num_bits, packed_dim=1, sym=False, )
slime/tools/convert_hf_to_int4_direct.py:30
↓ 1 callers
Function
pack_to_int32
( value, num_bits, packed_dim=1, sym=False, )
slime/slime/backends/megatron_utils/megatron_to_hf/processors/quantizer_compressed_tensors.py:135
↓ 1 callers
Method
pad
Returns id of padding token.
Megatron-LM/megatron/core/tokenizers/text/text_tokenizer.py:171
↓ 1 callers
Method
pad_encoder_prompts_tokens
Method to pad input prompts Given a list of prompts, pad them all to uniform length Args: encoder_prompts_tokens_list (L
Megatron-LM/megatron/core/inference/model_inference_wrappers/t5/t5_inference_wrapper.py:126
↓ 1 callers
Method
pad_input_prompt_tokens
Method to pad input prompts Given a list of prompts, pad them all to uniform length Args: batch_prompt_tokens_list (List
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:437
↓ 1 callers
Function
pad_neighbours_for_query_only
(args, nb_tokens, pad_id, ft_neighbours)
Megatron-LM/tools/retro/text_generation/retro_text_generation.py:57
↓ 1 callers
Function
pad_to_expected_shape
Pad tensor to the expected shape.
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:296
↓ 1 callers
Function
padded_forward
(input_tensor, *args, **kwargs)
Megatron-LM/megatron/core/fp8_utils.py:691
↓ 1 callers
Function
parquet_reader
(p)
slime/slime/utils/data.py:52
↓ 1 callers
Function
parse_args
()
terminal-rl/router_server.py:376
↓ 1 callers
Function
parse_args
()
terminal-rl/remote/pool_server.py:575
↓ 1 callers
Function
parse_args
()
gui-rl/env_pool_server.py:476
↓ 1 callers
Function
parse_args
()
swe-rl/server/swe_env_pool_server.py:341
↓ 1 callers
Function
parse_args
()
openclaw-tinker/run.py:25
↓ 1 callers
Function
parse_args
()
slime/tools/convert_hf_to_int4.py:13
↓ 1 callers
Function
parse_args
()
slime/tools/convert_k2_thinking_int4_to_bf16.py:158
↓ 1 callers
Function
parse_args
()
slime/tools/convert_hf_to_int4_direct.py:281
↓ 1 callers
Function
parse_args_train_backend
()
slime/slime/utils/arguments.py:1748
↓ 1 callers
Function
parse_failed_job
(logs: List[str])
Megatron-LM/tests/test_utils/python_scripts/launch_jet_workload.py:224
↓ 1 callers
Function
parse_fsdp_cli
(extra_args_provider=None)
slime/slime/backends/fsdp_utils/arguments.py:70
↓ 1 callers
Function
parse_img_path
(text)
Megatron-LM/examples/multimodal/evaluation/mmmu_utils.py:68
↓ 1 callers
Method
parse_megatron_args
Parse Megatron arguments by forcibly overwriting sys.argv. Populates self.margs and self.checkpoint_args.
Megatron-LM/tools/checkpoint/loader_base.py:37
↓ 1 callers
Method
parse_megatron_args
Parse Megatron arguments by forcibly overwriting sys.argv. Populates self.margs and self.checkpoint_args.
Megatron-LM/tools/checkpoint/saver_base.py:90
↓ 1 callers
Function
parse_qt_slider_colours
(colours_string)
gui-rl/desktop_env/evaluators/metrics/vlc.py:440
↓ 1 callers
Method
parse_response
(self, responses: list[InferenceResponse])
Megatron-LM/megatron/rl/inference/chat_templates.py:26
↓ 1 callers
Method
parse_response_to_openai_format
Parse sglang response to OpenAI compatible format Args: response: Raw response text from sglang Returns:
slime/examples/tau-bench/openai_tool_adapter.py:51
↓ 1 callers
Function
parse_tools
This function mimics the function call parser API from https://github.com/sgl-project/sglang/blob/main/python/sglang/srt/entrypoints/http_ser
slime/examples/tau-bench/sglang_tool_parser.py:7
↓ 1 callers
Function
parse_with_default_scheme
Ensure the URL has a scheme. If not, prepend 'http://' so it parses as host + path instead of just a path.
gui-rl/desktop_env/evaluators/metrics/utils.py:789
↓ 1 callers
Function
pass_at_k
Lower variance estimator of pass@k.
Megatron-LM/megatron/rl/agent/pass_at_evaluation_agent.py:14
↓ 1 callers
Function
patch_megatron_model
(model)
slime/slime/utils/megatron_bridge_utils.py:14
↓ 1 callers
Method
pause_submission
(self)
openclaw-rl/openclaw_rollout.py:65
↓ 1 callers
Method
pause_submission
(self)
openclaw-rl/oel/openclaw_oel_rollout.py:78
↓ 1 callers
Method
pause_submission
(self)
openclaw-opd/openclaw_opd_rollout.py:69
↓ 1 callers
Method
pause_submission
(self)
openclaw-combine/openclaw_combine_rollout.py:69
↓ 1 callers
Method
pause_submission
(self)
openclaw-combine/openclaw_combine_select_rollout.py:85
↓ 1 callers
Method
pause_submission
(self)
openclaw-tinker/rollout.py:80
↓ 1 callers
Method
periodic_reap
(self, interval: float = 60.0)
terminal-rl/remote/pool_server.py:317
↓ 1 callers
Function
pick_from_args
(args: Any, attrs: tuple[str, ...])
slime/slime/utils/eval_config.py:78
↓ 1 callers
Function
pixel_shuffle
Pixel shuffle based on InternVL but adapted for our use case. Args: x (torch.Tensor): Vision model outputs [num_tiles, img_seq_len, h_vis
Megatron-LM/megatron/core/models/multimodal/llava_model.py:1001
↓ 1 callers
Function
pooling
(pooler_output, last_hidden_state, attention_mask=None, pooling_method="mean")
slime/examples/search-r1/local_dense_retriever/retrieval_server.py:53
↓ 1 callers
Method
pop_backward
(self)
slime/slime/utils/routing_replay.py:33
↓ 1 callers
Method
pop_forward
(self)
slime/slime/utils/routing_replay.py:28
↓ 1 callers
Method
pop_tensors
Extracts the tensor data from the wrapped state dict, preserving metadata. Replaces the tensor data in sharded_tensors with device t
Megatron-LM/megatron/core/dist_checkpointing/tensor_aware_state_dict.py:193
↓ 1 callers
Method
post_forward_comm
Reduce scatter for SP after forward. This function is used to overlap shared experts with the dispatcher. It is only useful w
Megatron-LM/megatron/core/transformer/moe/shared_experts.py:239
↓ 1 callers
Function
post_language_model_processing
(lm_output, labels, logit_weights, parallel_output,
Megatron-LM/megatron/legacy/model/gpt_model.py:18
↓ 1 callers
Function
post_language_model_processing
(lm_output, pooled_output, lm_head, binary_head,
Megatron-LM/megatron/legacy/model/bert_model.py:96
↓ 1 callers
Method
post_process_requests
Handles post-processing for requests after a step. Args: request_ids (torch.Tensor): A list of request_ids f
Megatron-LM/megatron/core/inference/engines/dynamic_engine.py:767
↓ 1 callers
Function
post_training_step_callbacks
Run all post-training-step functions (e.g., FT heartbeats, GC).
Megatron-LM/megatron/training/training.py:1935
↓ 1 callers
Function
postprocess_hf_param
(args, megatron_param_name, hf_param_name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/__init__.py:15
↓ 1 callers
Function
postprocess_predictions
Extract action and content from prediction string
toolcall-rl/generate_with_retool.py:237
↓ 1 callers
Function
postprocess_predictions
(prediction: str)
slime/examples/search-r1/generate_with_search.py:111
↓ 1 callers
Function
postprocess_responses
Post-process response to ensure tag completeness
toolcall-rl/generate_with_retool.py:302
↓ 1 callers
Function
postprocess_responses
Post-process response to ensure tag completeness. Only used when SEARCH_R1_CONFIGS["return_logprob"] is False.
slime/examples/search-r1/generate_with_search.py:99
↓ 1 callers
Function
postprocess_sample_with_radix_tree
(args, sample: Sample, output: dict)
slime/slime/router/middleware_hub/radix_tree_middleware.py:158
↓ 1 callers
Method
pp_layers
(self, num_layers)
Megatron-LM/tests/unit_tests/export/trtllm/test_trtllm_single_device_converter.py:131
↓ 1 callers
Method
pre_forward_comm
All Gather for SP before forward. This function is used to overlap shared experts with the dispatcher. It is only useful when
Megatron-LM/megatron/core/transformer/moe/shared_experts.py:152
↓ 1 callers
Method
pre_reload_last_layer
Pre-reload the last layer of this chunk to hide reload latency.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:379
↓ 1 callers
Method
prep_inference_input
Preparing input data for inference, using respective wrapper's prep_inference_input method # pylint: disable=line-too-long Args:
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:1460
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen2.5_0.5B_gsm8k_short.py:11
↓ 1 callers
Function
prepare
Download model and convert checkpoint with MTP layers.
slime/tests/test_mimo_7B_mtp_only_grad.py:21
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_0.6B_fsdp_colocated_2xGPU.py:7
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_30B_A3B_r3.py:16
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_30B_A3B.py:16
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_4B_ppo.py:14
↓ 1 callers
Function
prepare
()
slime/tests/test_quick_start_glm4_9B.py:12
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen35_4b_fsdp.py:40
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen2.5_0.5B_gsm8k_async_short.py:11
↓ 1 callers
Function
prepare
()
slime/tests/test_external_rollout.py:15
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen2.5_0.5B_gsm8k.py:13
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_4B_fsdp_true_on_policy.py:14
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen2.5_0.5B_gsm8k_async.py:12
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_4B_ckpt.py:19
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_0.6B_parallel_check.py:14
↓ 1 callers
Function
prepare
()
slime/tests/test_moonlight_16B_A3B_r3.py:12
↓ 1 callers
Function
prepare
()
slime/tests/test_moonlight_16B_A3B.py:12
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_0.6B_megatron_fsdp_align.py:13
↓ 1 callers
Function
prepare
()
slime/tests/test_qwen3_0.6B_fsdp_distributed.py:10
← previous
next →
4,701–4,800 of 10,561, ranked by callers