My learning notes for ML SYS.
This report summarizes scanner evidence for zhaochenyang20/Awesome-ML-SYS-Tutorial, a HTML project with 7,455 GitHub stars. SynthScan examined 63,877 lines of code across 205 source files, recording 175 pattern matches distributed across 10 syntactic categories. The legacy adjusted pattern signal is 5.5 (Moderate pattern signal); it is not an authorship estimate.
Use the findings to ask whether maintainers can explain intent, verify dependencies, simplify unnecessary scaffolding, and focus review effort. Risk Payload v1 adds ownership, supply-chain, over-engineering, and review-effort vectors when available. These are prompts for human verification, not probabilities, security findings, or measured maintenance costs.
Use each vector to guide manual verification of the underlying findings.
Can maintainers explain intent, assumptions, and trade-offs? 114 finding(s).
Are imports declared and package/API names verified? 2 finding(s).
Are wrappers, guards, and repeated structures necessary? 119 finding(s).
Which changes need focused human review? 175 finding(s).
This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.
Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.
This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.
The scanner identified 175 distinct pattern matches across 10 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.
Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW⚡ | .claude/commands/learn-review.md | 16 | ### Step 1: 加载上下文 | COMMENT |
| LOW⚡ | .claude/commands/learn-review.md | 22 | ### Step 2: 分析输入 | COMMENT |
| LOW⚡ | .claude/commands/learn-review.md | 31 | ### Step 3: 内容检查 | COMMENT |
| LOW | .claude/commands/learn-review.md | 147 | ### Step 4: 输出检查报告 | COMMENT |
| LOW | .claude/commands/learn-review.md | 187 | ### Step 5: 翻译（需用户明确授权） | COMMENT |
| LOW⚡ | .claude/commands/learn-plan.md | 17 | ### Step 1: 加载上下文 | COMMENT |
| LOW⚡ | .claude/commands/learn-plan.md | 23 | ### Step 2: 分析输入 | COMMENT |
| LOW⚡ | .claude/commands/learn-plan.md | 33 | ### Step 3: 定位知识关系 | COMMENT |
| LOW⚡ | .claude/commands/learn-plan.md | 41 | ### Step 4: 广泛视野搜索 | COMMENT |
| LOW⚡ | .claude/commands/learn-plan.md | 48 | ### Step 5: 识别驱动问题 | COMMENT |
| LOW⚡ | .claude/commands/learn-plan.md | 55 | ### Step 6: 生成学习大纲 | COMMENT |
| LOW | .claude/commands/learn-plan.md | 140 | ### Step 7: 草稿分析（如有草稿输入） | COMMENT |
| LOW | .claude/commands/learn-plan.md | 158 | ### Step 8: 保存输出 | COMMENT |
| LOW | .claude/commands/learn-add.md | 13 | ### Step 1: 加载上下文 | COMMENT |
| LOW | .claude/commands/learn-add.md | 18 | ### Step 2: 解析输入 | COMMENT |
| LOW | .claude/commands/learn-add.md | 32 | ### Step 3: 提取元信息 | COMMENT |
| LOW⚡ | .claude/commands/learn-add.md | 72 | ### Step 4: 更新 referenced_by | COMMENT |
| LOW⚡ | .claude/commands/learn-add.md | 77 | ### Step 5: 更新系列信息 | COMMENT |
| LOW⚡ | .claude/commands/learn-add.md | 84 | ### Step 6: 确认并保存 | COMMENT |
| LOW | .claude/commands/learn-write.md | 32 | ### Step 1: 加载上下文 | COMMENT |
| LOW | .claude/commands/learn-write.md | 39 | ### Step 2: 分析输入 | COMMENT |
| LOW | .claude/commands/learn-write.md | 53 | ### Step 3: 源码与资源准备 | COMMENT |
| LOW | .claude/commands/learn-write.md | 64 | ### Step 4: 写作 | COMMENT |
| LOW | .claude/commands/learn-write.md | 160 | ### Step 5: 四维度自动检查 | COMMENT |
| LOW | .claude/commands/learn-write.md | 209 | ### Step 6: 保存输出 | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2-en.md | 183 | # Step 1: 暂时禁用 CUDA Graph | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2-en.md | 187 | # Step 2: 初始化 ModelWorker（此时不 capture graph） | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2-en.md | 190 | # Step 3: BF16 精度修正 | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2-en.md | 193 | # Step 4: 预分配 fast head 的 static KV cache | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2-en.md | 196 | # Step 5: 分配 persistent buffers + 挂载 audio decoder | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2-en.md | 199 | # Step 6: 此时 capture graph——包含完整的 forward + _decode_codebooks | COMMENT |
| LOW⚡ | torch/cuda-graph/readme_en.md | 73 | // Step 1: Create physical memory of the specified size on the specified GPU device | COMMENT |
| LOW⚡ | torch/cuda-graph/readme_en.md | 77 | // Step 2: Reserve a contiguous address range in the GPU's virtual address space | COMMENT |
| LOW⚡ | torch/cuda-graph/readme_en.md | 81 | // Step 3: Map physical memory to the previously reserved virtual address space | COMMENT |
| LOW | torch/cuda-graph/readme_en.md | 92 | // Step 4: Record memory allocation information in the metadata map | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2.md | 183 | # Step 1: 暂时禁用 CUDA Graph | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2.md | 187 | # Step 2: 初始化 ModelWorker（此时不 capture graph） | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2.md | 190 | # Step 3: BF16 精度修正 | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2.md | 193 | # Step 4: 预分配 fast head 的 static KV cache | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2.md | 196 | # Step 5: 分配 persistent buffers + 挂载 audio decoder | COMMENT |
| LOW⚡ | torch/cuda-graph/readme-2.md | 199 | # Step 6: 此时 capture graph——包含完整的 forward + _decode_codebooks | COMMENT |
| LOW | …fast_tokenization/verRL-multiturn-fast-tokenization.md | 24 | ### Step 1: Download GSM8K Dataset | COMMENT |
| LOW | …fast_tokenization/verRL-multiturn-fast-tokenization.md | 33 | ### Step 2: Validate Multi-Turn Rollout for Qwen2.5-3B | COMMENT |
| LOW | …fast_tokenization/verRL-multiturn-fast-tokenization.md | 49 | ### Step 3: Validate Multi-Turn Rollout for Qwen3-4B | COMMENT |
| LOW | sglang/light-duoattention/light-duoattention.md | 88 | # Step 3: Apply masking rules using direct, absolute positions | COMMENT |
| LOW | sglang/light-duoattention/light-duoattention.md | 115 | # Step 4: Apply the mask if needed | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | rlhf/slime/code-walk-through/readme.md | 0 | create placement groups for actor and rollout engines. | STRING |
| HIGH | rlhf/slime/code-walk-through/readme_en.md | 0 | create placement groups for actor and rollout engines. | STRING |
| HIGH | rlhf/slime/code-walk-through/original/part-2.md | 0 | create placement groups for actor and rollout engines. | STRING |
| HIGH | rlhf/slime/code-walk-through/readme.md | 0 | convert inference generated samples to training data. | STRING |
| HIGH | rlhf/slime/code-walk-through/readme_en.md | 0 | convert inference generated samples to training data. | STRING |
| HIGH | rlhf/slime/code-walk-through/original/part-4.md | 0 | convert inference generated samples to training data. | STRING |
| HIGH | rlhf/slime/code-walk-through/readme.md | 0 | an example to implement the generate_rollout function for an rule based rm rollout generation. args: args: the whole arg | STRING |
| HIGH | rlhf/slime/code-walk-through/readme_en.md | 0 | an example to implement the generate_rollout function for an rule based rm rollout generation. args: args: the whole arg | STRING |
| HIGH | rlhf/slime/code-walk-through/original/part-5.md | 0 | an example to implement the generate_rollout function for an rule based rm rollout generation. args: args: the whole arg | STRING |
| HIGH | rlhf/verl/multi-turn/tool_examples/profile_en.md | 0 | synchronized sglang rollout engine. args: actor_module: huggingface model name or path to the model. the model should be | STRING |
| HIGH | rlhf/verl/multi-turn/tool_examples/profile.md | 0 | synchronized sglang rollout engine. args: actor_module: huggingface model name or path to the model. the model should be | STRING |
| HIGH | rlhf/verl/multi-turn/code-walk-through/readme.md | 0 | synchronized sglang rollout engine. args: actor_module: huggingface model name or path to the model. the model should be | STRING |
| HIGH | rlhf/verl/multi-turn/code-walk-through/readme_EN.md | 0 | synchronized sglang rollout engine. args: actor_module: huggingface model name or path to the model. the model should be | STRING |
| HIGH | sglang/online-update-weights/readme.md | 0 | broadcast weight to all vllm workers from source rank 0 (actor model) | STRING |
| HIGH | sglang/sglang-worker/readme.md | 0 | broadcast weight to all vllm workers from source rank 0 (actor model) | STRING |
| HIGH | sglang/sglang-worker/readme-CN.md | 0 | broadcast weight to all vllm workers from source rank 0 (actor model) | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 63 | # --------------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 67 | # --------------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 69 | # --------------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 73 | # --------------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 75 | # --------------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 77 | # --------------------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 184 | # ------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 186 | # ------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 189 | # ------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 191 | # ------------------------------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 267 | # -------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 269 | # -------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 272 | # -------------------------------------------------------------- | COMMENT |
| MEDIUM⚡ | rlhf/OpenRLHF/readme.md | 274 | # -------------------------------------------------------------- | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | rlhf/OpenRLHF/readme.md | 109 | def get_advantages_and_returns(self, values, rewards, start): | CODE |
| LOW | rlhf/sys-design/readme-1.md | 16 | def _preprocess_tensor_for_update_weights(tensor: torch.Tensor): | CODE |
| LOW | rlhf/sys-design/readme-1.md | 72 | def _preprocess_tensor_for_update_weights(tensor: torch.Tensor): | CODE |
| LOW | rlhf/sys-design/readme-1.md | 213 | def update_weights_from_tensor( | CODE |
| LOW | rlhf/sys-design/readme-1.md | 280 | def update_weights_from_tensor(self): | CODE |
| LOW | rlhf/sys-design/readme-1.md | 376 | def _update_converted_params_from_tensor(self, converted_named_tensors): | CODE |
| LOW | rlhf/sys-design/readme-1-EN.md | 16 | def _preprocess_tensor_for_update_weights(tensor: torch.Tensor): | CODE |
| LOW | rlhf/sys-design/readme-1-EN.md | 71 | def _preprocess_tensor_for_update_weights(tensor: torch.Tensor): | CODE |
| LOW | rlhf/sys-design/readme-1-EN.md | 212 | def update_weights_from_tensor( | CODE |
| LOW | rlhf/sys-design/readme-1-EN.md | 279 | def update_weights_from_tensor(self): | CODE |
| LOW | rlhf/sys-design/readme-1-EN.md | 375 | def _update_converted_params_from_tensor(self, converted_named_tensors): | CODE |
| LOW | rlhf/slime/code-walk-through/readme.md | 193 | def _get_current_node_ip_and_free_port(start_port=10000, consecutive=1): | CODE |
| LOW | rlhf/slime/code-walk-through/readme.md | 804 | def _convert_samples_to_train_data(self, samples: Union[list[Sample], list[list[Sample]]]): | CODE |
| LOW | rlhf/slime/code-walk-through/readme.md | 900 | def generate_abortable_samples(args, rollout_id, data_source, evaluation=False): | CODE |
| LOW | rlhf/slime/code-walk-through/readme_en.md | 190 | def _get_current_node_ip_and_free_port(start_port=10000, consecutive=1): | CODE |
| LOW | rlhf/slime/code-walk-through/readme_en.md | 799 | def _convert_samples_to_train_data(self, samples: Union[list[Sample], list[list[Sample]]]): | CODE |
| LOW | rlhf/slime/code-walk-through/readme_en.md | 895 | def generate_abortable_samples(args, rollout_id, data_source, evaluation=False): | CODE |
| LOW | rlhf/slime/code-walk-through/original/part-4.md | 306 | def _convert_samples_to_train_data(self, samples: Union[list[Sample], list[list[Sample]]]): | CODE |
| LOW | rlhf/slime/code-walk-through/original/part-2.md | 153 | def _get_current_node_ip_and_free_port(start_port=10000, consecutive=1): | CODE |
| LOW | rlhf/verl/multi-turn/tool_examples/profile.md | 342 | async def process_request_with_monitoring(req): | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2-EN.md | 754 | def _req_level_generate_sequences(self, prompts: DataProto, **kwargs) -> DataProto: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2-EN.md | 1026 | def _req_level_generate_sequences(self, prompts: DataProto, **kwargs) -> DataProto: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2-EN.md | 1178 | def _preprocess_prompt_to_async_rollout_requests(self, prompts: DataProto, n: int) -> list[AsyncRolloutRequest]: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2-EN.md | 1404 | async def calc_reward_and_release_fn(name: str, tool: BaseTool): | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme.md | 983 | async def release_memory_occupation(self): | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme.md | 991 | async def update_weights_from_tensor( | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme_EN.md | 1048 | async def release_memory_occupation(self): | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme_EN.md | 1056 | async def update_weights_from_tensor( | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2.md | 766 | def _req_level_generate_sequences(self, prompts: DataProto, **kwargs) -> DataProto: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2.md | 1038 | def _req_level_generate_sequences(self, prompts: DataProto, **kwargs) -> DataProto: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2.md | 1190 | def _preprocess_prompt_to_async_rollout_requests(self, prompts: DataProto, n: int) -> list[AsyncRolloutRequest]: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2.md | 1415 | async def calc_reward_and_release_fn(name: str, tool: BaseTool): | CODE |
| LOW | rlhf/verl/multi-turn/release_log/over_sample.md | 96 | async def process_request_with_monitoring(req): | CODE |
| LOW | rlhf/verl/server-based/veRL-server-based-rollout.md | 35 | def launch_server_from_verl_engine( | CODE |
| LOW | rlhf/areal/code-walk-through_CN.md | 901 | def _broadcast_and_redistribute_batch( | CODE |
| LOW | rlhf/areal/code-walk-through_CN.md | 951 | def split_padded_tensor_dict_into_mb_list( | CODE |
| LOW | rlhf/areal/code-walk-through_EN.md | 958 | def _broadcast_and_redistribute_batch( | CODE |
| LOW | rlhf/areal/code-walk-through_EN.md | 1008 | def split_padded_tensor_dict_into_mb_list( | CODE |
| LOW | sglang/sglang-verl-engine/readme.md | 8 | def update_weights_from_tensor( | CODE |
| LOW | sglang/sglang-worker/readme.md | 81 | def determine_num_available_blocks(self) -> Tuple[int, int] # Determine available KV cache blocks | CODE |
| LOW | sglang/sglang-worker/readme.md | 82 | def get_cache_block_size_bytes(self) -> int # Get cache block size | CODE |
| LOW | sglang/sglang-worker/readme-CN.md | 82 | def determine_num_available_blocks(self) -> Tuple[int, int] # 确定可用的 KV cache 块数 | CODE |
| LOW | sglang/sglang-worker/readme-CN.md | 83 | def get_cache_block_size_bytes(self) -> int # 获取缓存块大小 | CODE |
| LOW | sglang/quantization/quantization_architecture_en.md | 204 | def process_weights_after_loading(self, layer: Module) -> None: | CODE |
| LOW | sglang/quantization/quantization_architecture.md | 204 | def process_weights_after_loading(self, layer: Module) -> None: | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 324 | def test_init_parameter_update_group(cls): | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 421 | def update_parameter_from_distributed(self, name, dtype, shape, empty_cache=False): | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 492 | def update_parameter_from_distributed(self, name, dtype, shape, empty_cache=False): | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 931 | def test_parameter_update_latency(self): | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 1012 | async def update_parameter_from_distributed( | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 323 | def test_init_parameter_update_group(cls): | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 421 | def update_parameter_from_distributed(self, name, dtype, shape, empty_cache=False): | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 492 | def update_parameter_from_distributed(self, name, dtype, shape, empty_cache=False): | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 932 | def test_parameter_update_latency(self): | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 1013 | async def update_parameter_from_distributed( | CODE |
| LOW | sglang/sglang-omni/README.md | 924 | def _inject_multimodal_embeds( | CODE |
| LOW | sglang/tiny-llm/day1.md | 12 | def scaled_dot_product_attention_simple( | CODE |
| LOW | sglang/dp-attention/readme.md | 172 | def launch_tensor_parallel_process( | CODE |
| LOW | sglang/scheduler-evolution/SGLang Scheduler 技术变迁.md | 52 | def process_batch_result_decode( | CODE |
| LOW | …lang/scheduler-evolution/SGLang Scheduler Evolution.md | 639 | def process_batch_result_decode( | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | torch/mem-snapshot/readme.md | 257 | except Exception: | CODE |
| LOW | torch/mem-snapshot/readme.md | 328 | except Exception as e: | CODE |
| LOW | torch/mem-snapshot/readme-en.md | 258 | except Exception: | CODE |
| LOW | torch/mem-snapshot/readme-en.md | 329 | except Exception as e: | CODE |
| LOW | rlhf/verl/multi-turn/tool_examples/profile.md | 369 | except Exception as e: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-6.md | 198 | except Exception: | CODE |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-6.md | 813 | except Exception as e: | CODE |
| LOW | rlhf/verl/multi-turn/release_log/over_sample.md | 138 | except Exception as e: | CODE |
| LOW | rlhf/verl/multi-turn/release_log/over_sample.md | 156 | except Exception as e: | CODE |
| LOW | rlhf/areal/code-walk-through_CN.md | 610 | except Exception as e: | CODE |
| LOW | rlhf/areal/code-walk-through_CN.md | 680 | except Exception as e: | CODE |
| LOW | rlhf/areal/code-walk-through_EN.md | 665 | except Exception as e: | CODE |
| LOW | rlhf/areal/code-walk-through_EN.md | 735 | except Exception as e: | CODE |
| LOW | sglang/online-update-weights/readme.md | 46 | except Exception as e: | CODE |
| LOW | sglang/online-update-weights/readme.md | 82 | except Exception as e: | CODE |
| LOW | sglang/online-update-weights/readme.md | 87 | except Exception as e: | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 356 | except Exception as e: | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 451 | except Exception as e: | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 543 | except Exception as e: | CODE |
| LOW | sglang/latency-accelerate-for-weight-updates/readme.md | 956 | except Exception as e: | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 355 | except Exception as e: | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 451 | except Exception as e: | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 543 | except Exception as e: | CODE |
| LOW | …ang/latency-accelerate-for-weight-updates/readme-CN.md | 957 | except Exception as e: | CODE |
| LOW⚡ | sglang/sglang-omni/README.md | 803 | except Exception as e: | CODE |
| LOW⚡ | sglang/sglang-omni/README.md | 808 | except Exception: | CODE |
| LOW⚡ | sglang/sglang-omni/README.md | 814 | except Exception: | CODE |
| LOW⚡ | sglang/sglang-omni/README.md | 821 | except Exception: | CODE |
| LOW | sglang/sglang-omni/README.md | 935 | except Exception: | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| CRITICAL | rlhf/OpenRLHF/develop-log.md | 632 | return self.llm.llm_engine.model_executor.driver_worker.init_process_group( | CODE |
| CRITICAL | rlhf/OpenRLHF/develop-log.md | 652 | return self.llm.llm_engine.model_executor.driver_worker.update_weight(name, dtype, shape, empty_cache) | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | rlhf/OpenRLHF/readme.md | 61 | kl_divergence_estimate = -self.kl_ctl * (log_probs - ref_log_probs) | COMMENT |
| LOW | rlhf/sys-design/readme-3.md | 241 | # b: 批次大小 (batch size) | COMMENT |
| LOW | rlhf/sys-design/readme-3.md | 301 | dim=3) | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2-EN.md | 689 | # TODO: implement actual tflpo and theoretical tflpo | COMMENT |
| LOW | rlhf/verl/multi-turn/code-walk-through/readme-2.md | 701 | # TODO: implement actual tflpo and theoretical tflpo | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | torch/torch-distributed/codes/send-recv-sync.py | 6 | CODE | |
| LOW | torch/torch-distributed/codes/send-recv-async.py | 7 | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | sglang/sglang-omni/cosmos3_omni_deep_dive.html | 668 | <text x="90" y="510" class="da" fill="#B87514">CommEngine / relay 是这一层的组件 —— 不是 ModelRunner 之上的 orchestration 层</t | CODE |