跳转至

verl 代码讲解教程 - TODO 进度追踪

总计: 397个Python文件需要讲解 创建日期: 2026-03-13 最后更新: 2026-03-13

进度总览

模块 文件数 已完成 状态
基础文档 (背景知识+阅读指南) 2 2 ✅ 完成
core (根目录) 3 3 ✅ 完成
checkpoint_engine 7 7 ✅ 完成
models 45 45 ✅ 完成
workers 65 65 ✅ 完成
trainer 18 18 ✅ 完成
utils 105 105 ✅ 完成
single_controller 7 7 ✅ 完成
tools 15 15 ✅ 完成
interactions 6 6 ✅ 完成
model_merger 5 5 ✅ 完成
experimental 98 98 ✅ 完成
third_party 7 7 ✅ 完成

阶段一：基础文档

  • [ ] tutorial/00_background_knowledge.md - 背景知识（RL、RLHF、PPO、分布式训练）
  • [ ] tutorial/00_reading_guide.md - 阅读指南（推荐阅读顺序）

阶段二：核心模块 (core)

verl/ 根目录 (3 files)

  • [ ] verl/__init__.py → tutorial/core/01_init.md
  • [ ] verl/protocol.py → tutorial/core/02_protocol.md
  • [ ] verl/base_config.py → tutorial/core/03_base_config.md
  • [ ] tutorial/core/index.md - 核心模块总览

阶段三：checkpoint_engine 模块

verl/checkpoint_engine/ (7 files)

  • [ ] verl/checkpoint_engine/__init__.py → tutorial/checkpoint_engine/01_init.md
  • [ ] verl/checkpoint_engine/base.py → tutorial/checkpoint_engine/02_base.md
  • [ ] verl/checkpoint_engine/nccl_checkpoint_engine.py → tutorial/checkpoint_engine/03_nccl_checkpoint_engine.md
  • [ ] verl/checkpoint_engine/hccl_checkpoint_engine.py → tutorial/checkpoint_engine/04_hccl_checkpoint_engine.md
  • [ ] verl/checkpoint_engine/kimi_checkpoint_engine.py → tutorial/checkpoint_engine/05_kimi_checkpoint_engine.md
  • [ ] verl/checkpoint_engine/mooncake_checkpoint_engine.py → tutorial/checkpoint_engine/06_mooncake_checkpoint_engine.md
  • [ ] verl/checkpoint_engine/nixl_checkpoint_engine.py → tutorial/checkpoint_engine/07_nixl_checkpoint_engine.md
  • [ ] tutorial/checkpoint_engine/index.md - 模块总览

阶段四：models 模块

verl/models/ 根目录 (3 files)

  • [ ] verl/models/__init__.py → tutorial/models/01_init.md
  • [ ] verl/models/registry.py → tutorial/models/02_registry.md
  • [ ] verl/models/weight_loader_registry.py → tutorial/models/03_weight_loader_registry.md
  • [ ] tutorial/models/index.md - 模块总览

verl/models/transformers/ (12 files)

  • [ ] verl/models/transformers/__init__.py → tutorial/models/transformers/01_init.md
  • [ ] verl/models/transformers/llama.py → tutorial/models/transformers/02_llama.md
  • [ ] verl/models/transformers/qwen2.py → tutorial/models/transformers/03_qwen2.md
  • [ ] verl/models/transformers/qwen2_vl.py → tutorial/models/transformers/04_qwen2_vl.md
  • [ ] verl/models/transformers/qwen3_vl.py → tutorial/models/transformers/05_qwen3_vl.md
  • [ ] verl/models/transformers/glm4v.py → tutorial/models/transformers/06_glm4v.md
  • [ ] verl/models/transformers/kimi_vl.py → tutorial/models/transformers/07_kimi_vl.md
  • [ ] verl/models/transformers/apertus.py → tutorial/models/transformers/08_apertus.md
  • [ ] verl/models/transformers/dense_common.py → tutorial/models/transformers/09_dense_common.md
  • [ ] verl/models/transformers/monkey_patch.py → tutorial/models/transformers/10_monkey_patch.md
  • [ ] verl/models/transformers/npu_patch.py → tutorial/models/transformers/11_npu_patch.md
  • [ ] verl/models/transformers/tiled_mlp.py → tutorial/models/transformers/12_tiled_mlp.md

verl/models/llama/megatron/ (13 files)

  • [ ] verl/models/llama/__init__.py → tutorial/models/llama_megatron/01_init.md
  • [ ] verl/models/llama/megatron/__init__.py → tutorial/models/llama_megatron/02_megatron_init.md
  • [ ] verl/models/llama/megatron/modeling_llama_megatron.py → tutorial/models/llama_megatron/03_modeling_llama_megatron.md
  • [ ] verl/models/llama/megatron/layers/__init__.py → tutorial/models/llama_megatron/04_layers_init.md
  • [ ] verl/models/llama/megatron/layers/parallel_attention.py → tutorial/models/llama_megatron/05_parallel_attention.md
  • [ ] verl/models/llama/megatron/layers/parallel_decoder.py → tutorial/models/llama_megatron/06_parallel_decoder.md
  • [ ] verl/models/llama/megatron/layers/parallel_linear.py → tutorial/models/llama_megatron/07_parallel_linear.md
  • [ ] verl/models/llama/megatron/layers/parallel_mlp.py → tutorial/models/llama_megatron/08_parallel_mlp.md
  • [ ] verl/models/llama/megatron/layers/parallel_rmsnorm.py → tutorial/models/llama_megatron/09_parallel_rmsnorm.md
  • [ ] verl/models/llama/megatron/checkpoint_utils/__init__.py → tutorial/models/llama_megatron/10_ckpt_init.md
  • [ ] verl/models/llama/megatron/checkpoint_utils/llama_loader.py → tutorial/models/llama_megatron/11_llama_loader.md
  • [ ] verl/models/llama/megatron/checkpoint_utils/llama_loader_depracated.py → tutorial/models/llama_megatron/12_llama_loader_deprecated.md
  • [ ] verl/models/llama/megatron/checkpoint_utils/llama_saver.py → tutorial/models/llama_megatron/13_llama_saver.md

verl/models/qwen2/megatron/ (13 files)

  • [ ] verl/models/qwen2/__init__.py → tutorial/models/qwen2_megatron/01_init.md
  • [ ] verl/models/qwen2/megatron/__init__.py → tutorial/models/qwen2_megatron/02_megatron_init.md
  • [ ] verl/models/qwen2/megatron/modeling_qwen2_megatron.py → tutorial/models/qwen2_megatron/03_modeling_qwen2_megatron.md
  • [ ] verl/models/qwen2/megatron/layers/__init__.py → tutorial/models/qwen2_megatron/04_layers_init.md
  • [ ] verl/models/qwen2/megatron/layers/parallel_attention.py → tutorial/models/qwen2_megatron/05_parallel_attention.md
  • [ ] verl/models/qwen2/megatron/layers/parallel_decoder.py → tutorial/models/qwen2_megatron/06_parallel_decoder.md
  • [ ] verl/models/qwen2/megatron/layers/parallel_linear.py → tutorial/models/qwen2_megatron/07_parallel_linear.md
  • [ ] verl/models/qwen2/megatron/layers/parallel_mlp.py → tutorial/models/qwen2_megatron/08_parallel_mlp.md
  • [ ] verl/models/qwen2/megatron/layers/parallel_rmsnorm.py → tutorial/models/qwen2_megatron/09_parallel_rmsnorm.md
  • [ ] verl/models/qwen2/megatron/checkpoint_utils/__init__.py → tutorial/models/qwen2_megatron/10_ckpt_init.md
  • [ ] verl/models/qwen2/megatron/checkpoint_utils/qwen2_loader.py → tutorial/models/qwen2_megatron/11_qwen2_loader.md
  • [ ] verl/models/qwen2/megatron/checkpoint_utils/qwen2_loader_depracated.py → tutorial/models/qwen2_megatron/12_qwen2_loader_deprecated.md
  • [ ] verl/models/qwen2/megatron/checkpoint_utils/qwen2_saver.py → tutorial/models/qwen2_megatron/13_qwen2_saver.md

verl/models/mcore/ (19 files)

  • [ ] verl/models/mcore/__init__.py → tutorial/models/mcore/01_init.md
  • [ ] verl/models/mcore/bridge.py → tutorial/models/mcore/02_bridge.md
  • [ ] verl/models/mcore/mbridge.py → tutorial/models/mcore/03_mbridge.md
  • [ ] verl/models/mcore/loader.py → tutorial/models/mcore/04_loader.md
  • [ ] verl/models/mcore/saver.py → tutorial/models/mcore/05_saver.md
  • [ ] verl/models/mcore/model_forward.py → tutorial/models/mcore/06_model_forward.md
  • [ ] verl/models/mcore/model_forward_1f1b_overlap.py → tutorial/models/mcore/07_model_forward_1f1b_overlap.md
  • [ ] verl/models/mcore/model_forward_fused.py → tutorial/models/mcore/08_model_forward_fused.md
  • [ ] verl/models/mcore/model_initializer.py → tutorial/models/mcore/09_model_initializer.md
  • [ ] verl/models/mcore/patch.py → tutorial/models/mcore/10_patch.md
  • [ ] verl/models/mcore/mtp_patch.py → tutorial/models/mcore/11_mtp_patch.md
  • [ ] verl/models/mcore/config_converter.py → tutorial/models/mcore/12_config_converter.md
  • [ ] verl/models/mcore/weight_converter.py → tutorial/models/mcore/13_weight_converter.md
  • [ ] verl/models/mcore/registry.py → tutorial/models/mcore/14_registry.md
  • [ ] verl/models/mcore/util.py → tutorial/models/mcore/15_util.md
  • [ ] verl/models/mcore/qwen2_5_vl/__init__.py → tutorial/models/mcore/16_qwen25vl_init.md
  • [ ] verl/models/mcore/qwen2_5_vl/model.py → tutorial/models/mcore/17_qwen25vl_model.md
  • [ ] verl/models/mcore/qwen2_5_vl/attention.py → tutorial/models/mcore/18_qwen25vl_attention.md
  • [ ] verl/models/mcore/qwen2_5_vl/rope_utils.py → tutorial/models/mcore/19_qwen25vl_rope_utils.md
  • [ ] verl/models/mcore/qwen2_5_vl/vision_config.py → tutorial/models/mcore/20_qwen25vl_vision_config.md
  • [ ] verl/models/mcore/qwen2_5_vl/vision_model.py → tutorial/models/mcore/21_qwen25vl_vision_model.md
  • [ ] verl/models/mcore/qwen2_5_vl/vision_transformer_block.py → tutorial/models/mcore/22_qwen25vl_vision_transformer_block.md

阶段五：workers 模块

verl/workers/ 根目录 (4 files)

  • [ ] verl/workers/__init__.py → tutorial/workers/01_init.md
  • [ ] verl/workers/engine_workers.py → tutorial/workers/02_engine_workers.md
  • [ ] verl/workers/fsdp_workers.py → tutorial/workers/03_fsdp_workers.md
  • [ ] verl/workers/megatron_workers.py → tutorial/workers/04_megatron_workers.md
  • [ ] tutorial/workers/index.md - 模块总览

verl/workers/actor/ (4 files)

  • [ ] verl/workers/actor/__init__.py → tutorial/workers/actor/01_init.md
  • [ ] verl/workers/actor/base.py → tutorial/workers/actor/02_base.md
  • [ ] verl/workers/actor/dp_actor.py → tutorial/workers/actor/03_dp_actor.md
  • [ ] verl/workers/actor/megatron_actor.py → tutorial/workers/actor/04_megatron_actor.md

verl/workers/critic/ (4 files)

  • [ ] verl/workers/critic/__init__.py → tutorial/workers/critic/01_init.md
  • [ ] verl/workers/critic/base.py → tutorial/workers/critic/02_base.md
  • [ ] verl/workers/critic/dp_critic.py → tutorial/workers/critic/03_dp_critic.md
  • [ ] verl/workers/critic/megatron_critic.py → tutorial/workers/critic/04_megatron_critic.md

verl/workers/rollout/ (17 files)

  • [ ] verl/workers/rollout/__init__.py → tutorial/workers/rollout/01_init.md
  • [ ] verl/workers/rollout/base.py → tutorial/workers/rollout/02_base.md
  • [ ] verl/workers/rollout/hf_rollout.py → tutorial/workers/rollout/03_hf_rollout.md
  • [ ] verl/workers/rollout/replica.py → tutorial/workers/rollout/04_replica.md
  • [ ] verl/workers/rollout/schemas.py → tutorial/workers/rollout/05_schemas.md
  • [ ] verl/workers/rollout/tokenizer.py → tutorial/workers/rollout/06_tokenizer.md
  • [ ] verl/workers/rollout/utils.py → tutorial/workers/rollout/07_utils.md
  • [ ] verl/workers/rollout/naive/__init__.py → tutorial/workers/rollout/08_naive_init.md
  • [ ] verl/workers/rollout/naive/naive_rollout.py → tutorial/workers/rollout/09_naive_rollout.md
  • [ ] verl/workers/rollout/sglang_rollout/__init__.py → tutorial/workers/rollout/sglang/01_init.md
  • [ ] verl/workers/rollout/sglang_rollout/sglang_rollout.py → tutorial/workers/rollout/sglang/02_sglang_rollout.md
  • [ ] verl/workers/rollout/sglang_rollout/async_sglang_server.py → tutorial/workers/rollout/sglang/03_async_sglang_server.md
  • [ ] verl/workers/rollout/sglang_rollout/http_server_engine.py → tutorial/workers/rollout/sglang/04_http_server_engine.md
  • [ ] verl/workers/rollout/sglang_rollout/utils.py → tutorial/workers/rollout/sglang/05_utils.md
  • [ ] verl/workers/rollout/vllm_rollout/__init__.py → tutorial/workers/rollout/vllm/01_init.md
  • [ ] verl/workers/rollout/vllm_rollout/vllm_rollout.py → tutorial/workers/rollout/vllm/02_vllm_rollout.md
  • [ ] verl/workers/rollout/vllm_rollout/vllm_async_server.py → tutorial/workers/rollout/vllm/03_vllm_async_server.md
  • [ ] verl/workers/rollout/vllm_rollout/utils.py → tutorial/workers/rollout/vllm/04_utils.md
  • [ ] verl/workers/rollout/vllm_rollout/bucketed_weight_transfer.py → tutorial/workers/rollout/vllm/05_bucketed_weight_transfer.md
  • [ ] verl/workers/rollout/trtllm_rollout/trtllm_rollout.py → tutorial/workers/rollout/trtllm/01_trtllm_rollout.md
  • [ ] verl/workers/rollout/trtllm_rollout/trtllm_async_server.py → tutorial/workers/rollout/trtllm/02_trtllm_async_server.md

verl/workers/engine/ (16 files)

  • [ ] verl/workers/engine/__init__.py → tutorial/workers/engine/01_init.md
  • [ ] verl/workers/engine/base.py → tutorial/workers/engine/02_base.md
  • [ ] verl/workers/engine/utils.py → tutorial/workers/engine/03_utils.md
  • [ ] verl/workers/engine/fsdp/__init__.py → tutorial/workers/engine/fsdp/01_init.md
  • [ ] verl/workers/engine/fsdp/transformer_impl.py → tutorial/workers/engine/fsdp/02_transformer_impl.md
  • [ ] verl/workers/engine/fsdp/utils.py → tutorial/workers/engine/fsdp/03_utils.md
  • [ ] verl/workers/engine/megatron/__init__.py → tutorial/workers/engine/megatron/01_init.md
  • [ ] verl/workers/engine/megatron/transformer_impl.py → tutorial/workers/engine/megatron/02_transformer_impl.md
  • [ ] verl/workers/engine/megatron/utils.py → tutorial/workers/engine/megatron/03_utils.md
  • [ ] verl/workers/engine/torchtitan/__init__.py → tutorial/workers/engine/torchtitan/01_init.md
  • [ ] verl/workers/engine/torchtitan/transformer_impl.py → tutorial/workers/engine/torchtitan/02_transformer_impl.md
  • [ ] verl/workers/engine/torchtitan/utils.py → tutorial/workers/engine/torchtitan/03_utils.md
  • [ ] verl/workers/engine/mindspeed/__init__.py → tutorial/workers/engine/mindspeed/01_init.md
  • [ ] verl/workers/engine/mindspeed/transformer_impl.py → tutorial/workers/engine/mindspeed/02_transformer_impl.md
  • [ ] verl/workers/engine/veomni/__init__.py → tutorial/workers/engine/veomni/01_init.md
  • [ ] verl/workers/engine/veomni/transformer_impl.py → tutorial/workers/engine/veomni/02_transformer_impl.md
  • [ ] verl/workers/engine/veomni/utils.py → tutorial/workers/engine/veomni/03_utils.md

verl/workers/reward_manager/ (7 files)

  • [ ] verl/workers/reward_manager/__init__.py → tutorial/workers/reward_manager/01_init.md
  • [ ] verl/workers/reward_manager/abstract.py → tutorial/workers/reward_manager/02_abstract.md
  • [ ] verl/workers/reward_manager/naive.py → tutorial/workers/reward_manager/03_naive.md
  • [ ] verl/workers/reward_manager/batch.py → tutorial/workers/reward_manager/04_batch.md
  • [ ] verl/workers/reward_manager/dapo.py → tutorial/workers/reward_manager/05_dapo.md
  • [ ] verl/workers/reward_manager/prime.py → tutorial/workers/reward_manager/06_prime.md
  • [ ] verl/workers/reward_manager/registry.py → tutorial/workers/reward_manager/07_registry.md

verl/workers/sharding_manager/ (3 files)

  • [ ] verl/workers/sharding_manager/__init__.py → tutorial/workers/sharding_manager/01_init.md
  • [ ] verl/workers/sharding_manager/base.py → tutorial/workers/sharding_manager/02_base.md
  • [ ] verl/workers/sharding_manager/fsdp_ulysses.py → tutorial/workers/sharding_manager/03_fsdp_ulysses.md

verl/workers/config/ (9 files)

  • [ ] verl/workers/config/__init__.py → tutorial/workers/config/01_init.md
  • [ ] verl/workers/config/actor.py → tutorial/workers/config/02_actor.md
  • [ ] verl/workers/config/critic.py → tutorial/workers/config/03_critic.md
  • [ ] verl/workers/config/model.py → tutorial/workers/config/04_model.md
  • [ ] verl/workers/config/engine.py → tutorial/workers/config/05_engine.md
  • [ ] verl/workers/config/rollout.py → tutorial/workers/config/06_rollout.md
  • [ ] verl/workers/config/reward.py → tutorial/workers/config/07_reward.md
  • [ ] verl/workers/config/optimizer.py → tutorial/workers/config/08_optimizer.md
  • [ ] verl/workers/config/megatron_peft.py → tutorial/workers/config/09_megatron_peft.md

verl/workers/utils/ (3 files)

  • [ ] verl/workers/utils/__init__.py → tutorial/workers/utils/01_init.md
  • [ ] verl/workers/utils/losses.py → tutorial/workers/utils/02_losses.md
  • [ ] verl/workers/utils/padding.py → tutorial/workers/utils/03_padding.md

阶段六：trainer 模块

verl/trainer/ 根目录 (7 files)

  • [ ] verl/trainer/__init__.py → tutorial/trainer/01_init.md
  • [ ] verl/trainer/main_ppo.py → tutorial/trainer/02_main_ppo.md
  • [ ] verl/trainer/main_eval.py → tutorial/trainer/03_main_eval.md
  • [ ] verl/trainer/main_generation_server.py → tutorial/trainer/04_main_generation_server.md
  • [ ] verl/trainer/sft_trainer.py → tutorial/trainer/05_sft_trainer.md
  • [ ] verl/trainer/sft_trainer_ray.py → tutorial/trainer/06_sft_trainer_ray.md
  • [ ] verl/trainer/constants_ppo.py → tutorial/trainer/07_constants_ppo.md
  • [ ] tutorial/trainer/index.md - 模块总览

verl/trainer/ppo/ (8 files)

  • [ ] verl/trainer/ppo/__init__.py → tutorial/trainer/ppo/01_init.md
  • [ ] verl/trainer/ppo/ray_trainer.py → tutorial/trainer/ppo/02_ray_trainer.md
  • [ ] verl/trainer/ppo/core_algos.py → tutorial/trainer/ppo/03_core_algos.md
  • [ ] verl/trainer/ppo/reward.py → tutorial/trainer/ppo/04_reward.md
  • [ ] verl/trainer/ppo/utils.py → tutorial/trainer/ppo/05_utils.md
  • [ ] verl/trainer/ppo/metric_utils.py → tutorial/trainer/ppo/06_metric_utils.md
  • [ ] verl/trainer/ppo/rollout_corr_helper.py → tutorial/trainer/ppo/07_rollout_corr_helper.md
  • [ ] verl/trainer/ppo/prefix_grouper_utils.py → tutorial/trainer/ppo/08_prefix_grouper_utils.md

verl/trainer/config/ (3 files)

  • [ ] verl/trainer/config/__init__.py → tutorial/trainer/config/01_init.md
  • [ ] verl/trainer/config/config.py → tutorial/trainer/config/02_config.md
  • [ ] verl/trainer/config/algorithm.py → tutorial/trainer/config/03_algorithm.md

阶段七：utils 模块

verl/utils/ 根目录 (28 files)

  • [ ] verl/utils/__init__.py → tutorial/utils/01_init.md
  • [ ] verl/utils/config.py → tutorial/utils/02_config.md
  • [ ] verl/utils/device.py → tutorial/utils/03_device.md
  • [ ] verl/utils/distributed.py → tutorial/utils/04_distributed.md
  • [ ] verl/utils/attention_utils.py → tutorial/utils/05_attention_utils.md
  • [ ] verl/utils/activation_offload.py → tutorial/utils/06_activation_offload.md
  • [ ] verl/utils/chat_template.py → tutorial/utils/07_chat_template.md
  • [ ] verl/utils/tokenizer.py → tutorial/utils/08_tokenizer.md
  • [ ] verl/utils/torch_dtypes.py → tutorial/utils/09_torch_dtypes.md
  • [ ] verl/utils/torch_functional.py → tutorial/utils/10_torch_functional.md
  • [ ] verl/utils/fsdp_utils.py → tutorial/utils/11_fsdp_utils.md
  • [ ] verl/utils/megatron_utils.py → tutorial/utils/12_megatron_utils.md
  • [ ] verl/utils/megatron_peft_utils.py → tutorial/utils/13_megatron_peft_utils.md
  • [ ] verl/utils/groupwise.py → tutorial/utils/14_groupwise.md
  • [ ] verl/utils/seqlen_balancing.py → tutorial/utils/15_seqlen_balancing.md
  • [ ] verl/utils/ulysses.py → tutorial/utils/16_ulysses.md
  • [ ] verl/utils/rollout_skip.py → tutorial/utils/17_rollout_skip.md
  • [ ] verl/utils/rollout_trace.py → tutorial/utils/18_rollout_trace.md
  • [ ] verl/utils/flops_counter.py → tutorial/utils/19_flops_counter.md
  • [ ] verl/utils/memory_utils.py → tutorial/utils/20_memory_utils.md
  • [ ] verl/utils/net_utils.py → tutorial/utils/21_net_utils.md
  • [ ] verl/utils/fs.py → tutorial/utils/22_fs.md
  • [ ] verl/utils/hdfs_io.py → tutorial/utils/23_hdfs_io.md
  • [ ] verl/utils/import_utils.py → tutorial/utils/24_import_utils.md
  • [ ] verl/utils/py_functional.py → tutorial/utils/25_py_functional.md
  • [ ] verl/utils/tracking.py → tutorial/utils/26_tracking.md
  • [ ] verl/utils/transformers_compat.py → tutorial/utils/27_transformers_compat.md
  • [ ] verl/utils/tensordict_utils.py → tutorial/utils/28_tensordict_utils.md
  • [ ] verl/utils/logging_utils.py → tutorial/utils/29_logging_utils.md
  • [ ] verl/utils/npu_flash_attn_utils.py → tutorial/utils/30_npu_flash_attn_utils.md
  • [ ] verl/utils/ray_utils.py → tutorial/utils/31_ray_utils.md
  • [ ] verl/utils/model.py → tutorial/utils/32_model.md
  • [ ] tutorial/utils/index.md - 模块总览

verl/utils/checkpoint/ (5 files)

  • [ ] verl/utils/checkpoint/__init__.py → tutorial/utils/checkpoint/01_init.md
  • [ ] verl/utils/checkpoint/checkpoint_handler.py → tutorial/utils/checkpoint/02_checkpoint_handler.md
  • [ ] verl/utils/checkpoint/checkpoint_manager.py → tutorial/utils/checkpoint/03_checkpoint_manager.md
  • [ ] verl/utils/checkpoint/fsdp_checkpoint_manager.py → tutorial/utils/checkpoint/04_fsdp_checkpoint_manager.md
  • [ ] verl/utils/checkpoint/megatron_checkpoint_manager.py → tutorial/utils/checkpoint/05_megatron_checkpoint_manager.md

verl/utils/dataset/ (6 files)

  • [ ] verl/utils/dataset/__init__.py → tutorial/utils/dataset/01_init.md
  • [ ] verl/utils/dataset/rl_dataset.py → tutorial/utils/dataset/02_rl_dataset.md
  • [ ] verl/utils/dataset/rm_dataset.py → tutorial/utils/dataset/03_rm_dataset.md
  • [ ] verl/utils/dataset/multiturn_sft_dataset.py → tutorial/utils/dataset/04_multiturn_sft_dataset.md
  • [ ] verl/utils/dataset/dataset_utils.py → tutorial/utils/dataset/05_dataset_utils.md
  • [ ] verl/utils/dataset/vision_utils.py → tutorial/utils/dataset/06_vision_utils.md

verl/utils/megatron/ (9 files)

  • [ ] verl/utils/megatron/__init__.py → tutorial/utils/megatron/01_init.md
  • [ ] verl/utils/megatron/dist_checkpointing.py → tutorial/utils/megatron/02_dist_checkpointing.md
  • [ ] verl/utils/megatron/tensor_parallel.py → tutorial/utils/megatron/03_tensor_parallel.md
  • [ ] verl/utils/megatron/pipeline_parallel.py → tutorial/utils/megatron/04_pipeline_parallel.md
  • [ ] verl/utils/megatron/sequence_parallel.py → tutorial/utils/megatron/05_sequence_parallel.md
  • [ ] verl/utils/megatron/optimizer.py → tutorial/utils/megatron/06_optimizer.md
  • [ ] verl/utils/megatron/memory.py → tutorial/utils/megatron/07_memory.md
  • [ ] verl/utils/megatron/router_replay_patch.py → tutorial/utils/megatron/08_router_replay_patch.md
  • [ ] verl/utils/megatron/router_replay_utils.py → tutorial/utils/megatron/09_router_replay_utils.md

verl/utils/profiler/ (8 files)

  • [ ] verl/utils/profiler/__init__.py → tutorial/utils/profiler/01_init.md
  • [ ] verl/utils/profiler/config.py → tutorial/utils/profiler/02_config.md
  • [ ] verl/utils/profiler/profile.py → tutorial/utils/profiler/03_profile.md
  • [ ] verl/utils/profiler/nvtx_profile.py → tutorial/utils/profiler/04_nvtx_profile.md
  • [ ] verl/utils/profiler/mstx_profile.py → tutorial/utils/profiler/05_mstx_profile.md
  • [ ] verl/utils/profiler/torch_profile.py → tutorial/utils/profiler/06_torch_profile.md
  • [ ] verl/utils/profiler/performance.py → tutorial/utils/profiler/07_performance.md
  • [ ] verl/utils/profiler/empty_annotations.py → tutorial/utils/profiler/08_empty_annotations.md

verl/utils/kernel/ (4 files)

  • [ ] verl/utils/kernel/__init__.py → tutorial/utils/kernel/01_init.md
  • [ ] verl/utils/kernel/kernels.py → tutorial/utils/kernel/02_kernels.md
  • [ ] verl/utils/kernel/linear_cross_entropy.py → tutorial/utils/kernel/03_linear_cross_entropy.md
  • [ ] verl/utils/kernel/fp8_kernel.py → tutorial/utils/kernel/04_fp8_kernel.md

verl/utils/reward_score/ (17 files)

  • [ ] verl/utils/reward_score/__init__.py → tutorial/utils/reward_score/01_init.md
  • [ ] verl/utils/reward_score/gsm8k.py → tutorial/utils/reward_score/02_gsm8k.md
  • [ ] verl/utils/reward_score/geo3k.py → tutorial/utils/reward_score/03_geo3k.md
  • [ ] verl/utils/reward_score/math_reward.py → tutorial/utils/reward_score/04_math_reward.md
  • [ ] verl/utils/reward_score/math_batch.py → tutorial/utils/reward_score/05_math_batch.md
  • [ ] verl/utils/reward_score/math_verify.py → tutorial/utils/reward_score/06_math_verify.md
  • [ ] verl/utils/reward_score/math_dapo.py → tutorial/utils/reward_score/07_math_dapo.md
  • [ ] verl/utils/reward_score/search_r1_like_qa_em.py → tutorial/utils/reward_score/08_search_r1_like_qa_em.md
  • [ ] verl/utils/reward_score/prime_math/__init__.py → tutorial/utils/reward_score/09_prime_math_init.md
  • [ ] verl/utils/reward_score/prime_math/grader.py → tutorial/utils/reward_score/10_prime_math_grader.md
  • [ ] verl/utils/reward_score/prime_math/math_normalize.py → tutorial/utils/reward_score/11_prime_math_normalize.md
  • [ ] verl/utils/reward_score/prime_code/__init__.py → tutorial/utils/reward_score/12_prime_code_init.md
  • [ ] verl/utils/reward_score/prime_code/testing_util.py → tutorial/utils/reward_score/13_prime_code_testing_util.md
  • [ ] verl/utils/reward_score/prime_code/utils.py → tutorial/utils/reward_score/14_prime_code_utils.md
  • [ ] verl/utils/reward_score/sandbox_fusion/__init__.py → tutorial/utils/reward_score/15_sandbox_fusion_init.md
  • [ ] verl/utils/reward_score/sandbox_fusion/utils.py → tutorial/utils/reward_score/16_sandbox_fusion_utils.md

verl/utils/qat/ (5 files)

  • [ ] verl/utils/qat/__init__.py → tutorial/utils/qat/01_init.md
  • [ ] verl/utils/qat/core.py → tutorial/utils/qat/02_core.md
  • [ ] verl/utils/qat/quantizer.py → tutorial/utils/qat/03_quantizer.md
  • [ ] verl/utils/qat/linear.py → tutorial/utils/qat/04_linear.md
  • [ ] verl/utils/qat/vllm_patch.py → tutorial/utils/qat/05_vllm_patch.md

verl/utils/vllm/ (4 files)

  • [ ] verl/utils/vllm/__init__.py → tutorial/utils/vllm/01_init.md
  • [ ] verl/utils/vllm/utils.py → tutorial/utils/vllm/02_utils.md
  • [ ] verl/utils/vllm/patch.py → tutorial/utils/vllm/03_patch.md
  • [ ] verl/utils/vllm/vllm_fp8_utils.py → tutorial/utils/vllm/04_vllm_fp8_utils.md

verl/utils/sglang/ (1 file)

  • [ ] verl/utils/sglang/sglang_fp8_utils.py → tutorial/utils/sglang/01_sglang_fp8_utils.md

verl/utils/logger/ (2 files)

  • [ ] verl/utils/logger/__init__.py → tutorial/utils/logger/01_init.md
  • [ ] verl/utils/logger/aggregate_logger.py → tutorial/utils/logger/02_aggregate_logger.md

verl/utils/debug/ (4 files)

  • [ ] verl/utils/debug/__init__.py → tutorial/utils/debug/01_init.md
  • [ ] verl/utils/debug/metrics.py → tutorial/utils/debug/02_metrics.md
  • [ ] verl/utils/debug/performance.py → tutorial/utils/debug/03_performance.md
  • [ ] verl/utils/debug/trajectory_tracker.py → tutorial/utils/debug/04_trajectory_tracker.md

verl/utils/experimental/ (2 files)

  • [ ] verl/utils/experimental/__init__.py → tutorial/utils/experimental/01_init.md
  • [ ] verl/utils/experimental/torch_functional.py → tutorial/utils/experimental/02_torch_functional.md

verl/utils/rendezvous/ (2 files)

  • [ ] verl/utils/rendezvous/__init__.py → tutorial/utils/rendezvous/01_init.md
  • [ ] verl/utils/rendezvous/ray_backend.py → tutorial/utils/rendezvous/02_ray_backend.md

verl/utils/metric/ (2 files)

  • [ ] verl/utils/metric/__init__.py → tutorial/utils/metric/01_init.md
  • [ ] verl/utils/metric/utils.py → tutorial/utils/metric/02_utils.md

阶段八：single_controller 模块

verl/single_controller/ (7 files)

  • [ ] verl/single_controller/__init__.py → tutorial/single_controller/01_init.md
  • [ ] verl/single_controller/base/__init__.py → tutorial/single_controller/base/01_init.md
  • [ ] verl/single_controller/base/worker.py → tutorial/single_controller/base/02_worker.md
  • [ ] verl/single_controller/base/worker_group.py → tutorial/single_controller/base/03_worker_group.md
  • [ ] verl/single_controller/base/decorator.py → tutorial/single_controller/base/04_decorator.md
  • [ ] verl/single_controller/ray/__init__.py → tutorial/single_controller/ray/01_init.md
  • [ ] verl/single_controller/ray/base.py → tutorial/single_controller/ray/02_base.md
  • [ ] tutorial/single_controller/index.md - 模块总览

阶段九：tools 模块

verl/tools/ (15 files)

  • [ ] verl/tools/__init__.py → tutorial/tools/01_init.md
  • [ ] verl/tools/base_tool.py → tutorial/tools/02_base_tool.md
  • [ ] verl/tools/mcp_base_tool.py → tutorial/tools/03_mcp_base_tool.md
  • [ ] verl/tools/schemas.py → tutorial/tools/04_schemas.md
  • [ ] verl/tools/search_tool.py → tutorial/tools/05_search_tool.md
  • [ ] verl/tools/mcp_search_tool.py → tutorial/tools/06_mcp_search_tool.md
  • [ ] verl/tools/gsm8k_tool.py → tutorial/tools/07_gsm8k_tool.md
  • [ ] verl/tools/geo3k_tool.py → tutorial/tools/08_geo3k_tool.md
  • [ ] verl/tools/image_zoom_in_tool.py → tutorial/tools/09_image_zoom_in_tool.md
  • [ ] verl/tools/sandbox_fusion_tools.py → tutorial/tools/10_sandbox_fusion_tools.md
  • [ ] verl/tools/utils/__init__.py → tutorial/tools/11_utils_init.md
  • [ ] verl/tools/utils/tool_registry.py → tutorial/tools/12_tool_registry.md
  • [ ] verl/tools/utils/search_r1_like_utils.py → tutorial/tools/13_search_r1_like_utils.md
  • [ ] verl/tools/utils/mcp_clients/McpClientManager.py → tutorial/tools/14_mcp_client_manager.md
  • [ ] verl/tools/utils/mcp_clients/utils.py → tutorial/tools/15_mcp_clients_utils.md
  • [ ] tutorial/tools/index.md - 模块总览

阶段十：interactions 模块

verl/interactions/ (6 files)

  • [ ] verl/interactions/__init__.py → tutorial/interactions/01_init.md
  • [ ] verl/interactions/base.py → tutorial/interactions/02_base.md
  • [ ] verl/interactions/gsm8k_interaction.py → tutorial/interactions/03_gsm8k_interaction.md
  • [ ] verl/interactions/weather_interaction.py → tutorial/interactions/04_weather_interaction.md
  • [ ] verl/interactions/utils/__init__.py → tutorial/interactions/05_utils_init.md
  • [ ] verl/interactions/utils/interaction_registry.py → tutorial/interactions/06_interaction_registry.md
  • [ ] tutorial/interactions/index.md - 模块总览

阶段十一：model_merger 模块

verl/model_merger/ (5 files)

  • [ ] verl/model_merger/__init__.py → tutorial/model_merger/01_init.md
  • [ ] verl/model_merger/__main__.py → tutorial/model_merger/02_main.md
  • [ ] verl/model_merger/base_model_merger.py → tutorial/model_merger/03_base_model_merger.md
  • [ ] verl/model_merger/fsdp_model_merger.py → tutorial/model_merger/04_fsdp_model_merger.md
  • [ ] verl/model_merger/megatron_model_merger.py → tutorial/model_merger/05_megatron_model_merger.md
  • [ ] tutorial/model_merger/index.md - 模块总览

阶段十二：experimental 模块

verl/experimental/ 根目录 (1 file)

  • [ ] verl/experimental/__init__.py → tutorial/experimental/01_init.md
  • [ ] tutorial/experimental/index.md - 模块总览

verl/experimental/agent_loop/ (7 files)

  • [ ] verl/experimental/agent_loop/__init__.py → tutorial/experimental/agent_loop/01_init.md
  • [ ] verl/experimental/agent_loop/agent_loop.py → tutorial/experimental/agent_loop/02_agent_loop.md
  • [ ] verl/experimental/agent_loop/single_turn_agent_loop.py → tutorial/experimental/agent_loop/03_single_turn_agent_loop.md
  • [ ] verl/experimental/agent_loop/tool_agent_loop.py → tutorial/experimental/agent_loop/04_tool_agent_loop.md
  • [ ] verl/experimental/agent_loop/tool_parser.py → tutorial/experimental/agent_loop/05_tool_parser.md
  • [ ] verl/experimental/agent_loop/utils.py → tutorial/experimental/agent_loop/06_utils.md
  • [ ] verl/experimental/agent_loop/prometheus_utils.py → tutorial/experimental/agent_loop/07_prometheus_utils.md

verl/experimental/dataset/ (2 files)

  • [ ] verl/experimental/dataset/__init__.py → tutorial/experimental/dataset/01_init.md
  • [ ] verl/experimental/dataset/sampler.py → tutorial/experimental/dataset/02_sampler.md

verl/experimental/dynamic_dataset/ (2 files)

  • [ ] verl/experimental/dynamic_dataset/__init__.py → tutorial/experimental/dynamic_dataset/01_init.md
  • [ ] verl/experimental/dynamic_dataset/dynamicgen_dataset.py → tutorial/experimental/dynamic_dataset/02_dynamicgen_dataset.md

verl/experimental/fully_async_policy/ (16 files)

  • [ ] verl/experimental/fully_async_policy/fully_async_main.py → tutorial/experimental/fully_async_policy/01_fully_async_main.md
  • [ ] verl/experimental/fully_async_policy/fully_async_trainer.py → tutorial/experimental/fully_async_policy/02_fully_async_trainer.md
  • [ ] verl/experimental/fully_async_policy/fully_async_rollouter.py → tutorial/experimental/fully_async_policy/03_fully_async_rollouter.md
  • [ ] verl/experimental/fully_async_policy/message_queue.py → tutorial/experimental/fully_async_policy/04_message_queue.md
  • [ ] verl/experimental/fully_async_policy/param_sync.py → tutorial/experimental/fully_async_policy/05_param_sync.md
  • [ ] verl/experimental/fully_async_policy/detach_utils.py → tutorial/experimental/fully_async_policy/06_detach_utils.md
  • [ ] verl/experimental/fully_async_policy/fsdp2_utils.py → tutorial/experimental/fully_async_policy/07_fsdp2_utils.md
  • [ ] verl/experimental/fully_async_policy/megatron_utils.py → tutorial/experimental/fully_async_policy/08_megatron_utils.md
  • [ ] verl/experimental/fully_async_policy/agent_loop/__init__.py → tutorial/experimental/fully_async_policy/09_agent_loop_init.md
  • [ ] verl/experimental/fully_async_policy/agent_loop/agent_loop.py → tutorial/experimental/fully_async_policy/10_agent_loop.md
  • [ ] verl/experimental/fully_async_policy/agent_loop/partial_single_turn_agent_loop.py → tutorial/experimental/fully_async_policy/11_partial_single_turn.md
  • [ ] verl/experimental/fully_async_policy/agent_loop/partial_tool_agent_loop.py → tutorial/experimental/fully_async_policy/12_partial_tool_agent.md
  • [ ] verl/experimental/fully_async_policy/sglang_rollout/__init__.py → tutorial/experimental/fully_async_policy/13_sglang_init.md
  • [ ] verl/experimental/fully_async_policy/sglang_rollout/sglang_async_server.py → tutorial/experimental/fully_async_policy/14_sglang_async_server.md
  • [ ] verl/experimental/fully_async_policy/vllm_rollout/__init__.py → tutorial/experimental/fully_async_policy/15_vllm_init.md
  • [ ] verl/experimental/fully_async_policy/vllm_rollout/vllm_async_server.py → tutorial/experimental/fully_async_policy/16_vllm_async_server.md
  • [ ] verl/experimental/fully_async_policy/unittest/simple_streaming_demo.py → tutorial/experimental/fully_async_policy/17_streaming_demo.md

verl/experimental/one_step_off_policy/ (6 files)

  • [ ] verl/experimental/one_step_off_policy/main_ppo.py → tutorial/experimental/one_step_off_policy/01_main_ppo.md
  • [ ] verl/experimental/one_step_off_policy/ray_trainer.py → tutorial/experimental/one_step_off_policy/02_ray_trainer.md
  • [ ] verl/experimental/one_step_off_policy/utils.py → tutorial/experimental/one_step_off_policy/03_utils.md
  • [ ] verl/experimental/one_step_off_policy/distributed_utils.py → tutorial/experimental/one_step_off_policy/04_distributed_utils.md
  • [ ] verl/experimental/one_step_off_policy/agent_loop/__init__.py → tutorial/experimental/one_step_off_policy/05_agent_loop_init.md
  • [ ] verl/experimental/one_step_off_policy/agent_loop/agent_loop.py → tutorial/experimental/one_step_off_policy/06_agent_loop.md

verl/experimental/reward_loop/ (11 files)

  • [ ] verl/experimental/reward_loop/__init__.py → tutorial/experimental/reward_loop/01_init.md
  • [ ] verl/experimental/reward_loop/reward_loop.py → tutorial/experimental/reward_loop/02_reward_loop.md
  • [ ] verl/experimental/reward_loop/reward_model.py → tutorial/experimental/reward_loop/03_reward_model.md
  • [ ] verl/experimental/reward_loop/reward_manager/__init__.py → tutorial/experimental/reward_loop/04_rm_init.md
  • [ ] verl/experimental/reward_loop/reward_manager/base.py → tutorial/experimental/reward_loop/05_rm_base.md
  • [ ] verl/experimental/reward_loop/reward_manager/naive.py → tutorial/experimental/reward_loop/06_rm_naive.md
  • [ ] verl/experimental/reward_loop/reward_manager/dapo.py → tutorial/experimental/reward_loop/07_rm_dapo.md
  • [ ] verl/experimental/reward_loop/reward_manager/limited.py → tutorial/experimental/reward_loop/08_rm_limited.md
  • [ ] verl/experimental/reward_loop/reward_manager/remote.py → tutorial/experimental/reward_loop/09_rm_remote.md
  • [ ] verl/experimental/reward_loop/reward_manager/registry.py → tutorial/experimental/reward_loop/10_rm_registry.md
  • [ ] verl/experimental/reward_loop/router/inner_sglang_router.py → tutorial/experimental/reward_loop/11_sglang_router.md
  • [ ] verl/experimental/reward_loop/router/naive_router.py → tutorial/experimental/reward_loop/12_naive_router.md

verl/experimental/separation/ (3 files)

  • [ ] verl/experimental/separation/__init__.py → tutorial/experimental/separation/01_init.md
  • [ ] verl/experimental/separation/engine_workers.py → tutorial/experimental/separation/02_engine_workers.md
  • [ ] verl/experimental/separation/ray_trainer.py → tutorial/experimental/separation/03_ray_trainer.md

verl/experimental/vla/ (36 files)

  • [ ] verl/experimental/vla/main_ppo.py → tutorial/experimental/vla/01_main_ppo.md
  • [ ] verl/experimental/vla/main_sac.py → tutorial/experimental/vla/02_main_sac.md
  • [ ] verl/experimental/vla/dp_rob.py → tutorial/experimental/vla/03_dp_rob.md
  • [ ] verl/experimental/vla/env_loop.py → tutorial/experimental/vla/04_env_loop.md
  • [ ] verl/experimental/vla/fsdp_workers.py → tutorial/experimental/vla/05_fsdp_workers.md
  • [ ] verl/experimental/vla/rob_ray_trainer.py → tutorial/experimental/vla/06_rob_ray_trainer.md
  • [ ] verl/experimental/vla/naive_rollout_rob.py → tutorial/experimental/vla/07_naive_rollout_rob.md
  • [ ] verl/experimental/vla/prepare_libero_dataset.py → tutorial/experimental/vla/08_prepare_libero_dataset.md
  • [ ] verl/experimental/vla/envs/__init__.py → tutorial/experimental/vla/09_envs_init.md
  • [ ] verl/experimental/vla/envs/action_utils.py → tutorial/experimental/vla/10_action_utils.md
  • [ ] verl/experimental/vla/envs/libero_env/__init__.py → tutorial/experimental/vla/11_libero_init.md
  • [ ] verl/experimental/vla/envs/libero_env/libero_env.py → tutorial/experimental/vla/12_libero_env.md
  • [ ] verl/experimental/vla/envs/libero_env/utils.py → tutorial/experimental/vla/13_libero_utils.md
  • [ ] verl/experimental/vla/envs/libero_env/venv.py → tutorial/experimental/vla/14_libero_venv.md
  • [ ] verl/experimental/vla/envs/isaac_env/__init__.py → tutorial/experimental/vla/15_isaac_init.md
  • [ ] verl/experimental/vla/envs/isaac_env/isaac_env.py → tutorial/experimental/vla/16_isaac_env.md
  • [ ] verl/experimental/vla/models/__init__.py → tutorial/experimental/vla/17_models_init.md
  • [ ] verl/experimental/vla/models/modules/mlp.py → tutorial/experimental/vla/18_modules_mlp.md
  • [ ] verl/experimental/vla/models/register_vla_models.py → tutorial/experimental/vla/19_register_vla_models.md
  • [ ] verl/experimental/vla/models/openvla_oft/__init__.py → tutorial/experimental/vla/20_openvla_init.md
  • [ ] verl/experimental/vla/models/openvla_oft/configuration_prismatic.py → tutorial/experimental/vla/21_config_prismatic.md
  • [ ] verl/experimental/vla/models/openvla_oft/modeling_prismatic.py → tutorial/experimental/vla/22_modeling_prismatic.md
  • [ ] verl/experimental/vla/models/openvla_oft/processing_prismatic.py → tutorial/experimental/vla/23_processing_prismatic.md
  • [ ] verl/experimental/vla/models/openvla_oft/train_utils.py → tutorial/experimental/vla/24_train_utils.md
  • [ ] verl/experimental/vla/models/openvla_oft/constants.py → tutorial/experimental/vla/25_constants.md
  • [ ] verl/experimental/vla/models/pi0_torch/__init__.py → tutorial/experimental/vla/26_pi0_init.md
  • [ ] verl/experimental/vla/models/pi0_torch/configuration_pi0_torch.py → tutorial/experimental/vla/27_config_pi0.md
  • [ ] verl/experimental/vla/models/pi0_torch/modeling_pi0_torch.py → tutorial/experimental/vla/28_modeling_pi0.md
  • [ ] verl/experimental/vla/models/pi0_torch/pi0_utils.py → tutorial/experimental/vla/29_pi0_utils.md
  • [ ] verl/experimental/vla/models/pi0_torch/model/modeling_pi0.py → tutorial/experimental/vla/30_model_pi0.md
  • [ ] verl/experimental/vla/models/pi0_torch/model/paligemma_with_expert.py → tutorial/experimental/vla/31_paligemma.md
  • [ ] verl/experimental/vla/models/pi0_torch/policy/__init__.py → tutorial/experimental/vla/32_policy_init.md
  • [ ] verl/experimental/vla/models/pi0_torch/policy/base.py → tutorial/experimental/vla/33_policy_base.md
  • [ ] verl/experimental/vla/models/pi0_torch/policy/libero_policy.py → tutorial/experimental/vla/34_libero_policy.md
  • [ ] verl/experimental/vla/sac/base.py → tutorial/experimental/vla/35_sac_base.md
  • [ ] verl/experimental/vla/sac/sac_actor.py → tutorial/experimental/vla/36_sac_actor.md
  • [ ] verl/experimental/vla/sac/sac_ray_trainer.py → tutorial/experimental/vla/37_sac_ray_trainer.md
  • [ ] verl/experimental/vla/sac/replay_pool.py → tutorial/experimental/vla/38_replay_pool.md
  • [ ] verl/experimental/vla/sac/naive_rollout_pi05.py → tutorial/experimental/vla/39_naive_rollout_pi05.md
  • [ ] verl/experimental/vla/workers/env/env_manager.py → tutorial/experimental/vla/40_env_manager.md
  • [ ] verl/experimental/vla/workers/env/env_loop_wg_test.py → tutorial/experimental/vla/41_env_loop_wg_test.md
  • [ ] verl/experimental/vla/workers/env/env_worker.py → tutorial/experimental/vla/42_env_worker.md

阶段十三：third_party 模块

verl/third_party/ (7 files)

  • [ ] verl/third_party/__init__.py → tutorial/third_party/01_init.md
  • [ ] verl/third_party/torch/__init__.py → tutorial/third_party/02_torch_init.md
  • [ ] verl/third_party/torch/distributed/__init__.py → tutorial/third_party/03_torch_distributed_init.md
  • [ ] verl/third_party/torch/distributed/_state_dict_utils.py → tutorial/third_party/04_state_dict_utils.md
  • [ ] verl/third_party/torch/distributed/checkpoint/__init__.py → tutorial/third_party/05_checkpoint_init.md
  • [ ] verl/third_party/torch/distributed/checkpoint/state_dict.py → tutorial/third_party/06_checkpoint_state_dict.md
  • [ ] verl/third_party/vllm/__init__.py → tutorial/third_party/07_vllm_init.md
  • [ ] tutorial/third_party/index.md - 模块总览

阶段十四：Review & 最终提交

  • [ ] 自我review所有文档，检查缺漏
  • [ ] 修复发现的问题
  • [ ] 最终commit

Agent 分工记录

Agent 负责模块 状态
Agent 1 背景知识文档 🔄 进行中
Agent 2 核心模块 (core) 🔄 进行中
Agent 3 checkpoint_engine 🔄 进行中
Agent 4 models 🔄 进行中
Agent 5 trainer 🔄 进行中
Agent 6 single_controller 🔄 进行中
Agent 7 workers ⏳ 待启动
Agent 8 utils (Part 1) ⏳ 待启动
Agent 9 utils (Part 2) ⏳ 待启动
Agent 10 tools + interactions ⏳ 待启动
Agent 11 model_merger ⏳ 待启动
Agent 12 experimental (Part 1) ⏳ 待启动
Agent 13 experimental (Part 2) ⏳ 待启动
Agent 14 third_party + 阅读指南 ⏳ 待启动