[v0.5.10][5] Disable piecewise cuda graph to avoid NVLS oom#935
Merged
yueming-yuan merged 38 commits intomainfrom Apr 9, 2026
Merged
[v0.5.10][5] Disable piecewise cuda graph to avoid NVLS oom#935yueming-yuan merged 38 commits intomainfrom
yueming-yuan merged 38 commits intomainfrom
Conversation
Contributor
There was a problem hiding this comment.
Code Review
This pull request addresses NCCL 'out of memory' errors related to NVLS multicast handles by introducing the --sglang-enforce-disable-flashinfer-allreduce-fusion flag in documentation and E2E tests. Feedback suggests adding the specific SGLang version (v0.5.10) to the documentation for clarity and ensuring a trailing newline is present in the markdown file.
|
|
||
| 14. **NCCL error: `Failed to bind NVLink SHARP (NVLS) Multicast memory ... CUDA error 2 'out of memory'` in colocate mode.** | ||
|
|
||
| This happens when SGLang's flashinfer allreduce fusion allocates NVLS (NVLink SHARP) multicast handles on the NVSwitch, leaving insufficient handles for Megatron's training NCCL. NVSwitch has a hardware-limited number of multicast handle slots. In colocate mode, SGLang and Megatron share the same GPUs, and SGLang's NVLS handles are not released during sleep. |
Contributor
There was a problem hiding this comment.
The explanation mentions that this issue happens when SGLang's flashinfer allreduce fusion allocates NVLS handles. It would be beneficial to specify that this behavior was introduced in SGLang v0.5.10 (switching from IPC to NVLS), as users on older versions might not encounter this or might find the flag missing.
|
|
||
| Fix: add `--sglang-enforce-disable-flashinfer-allreduce-fusion` to your training command. This disables flashinfer's NVLS-based allreduce fusion in SGLang, reserving NVLS handles for Megatron training. | ||
|
|
||
| This issue has been observed on H100 with Qwen3-30B-A3B in colocate mode. No newline at end of file |
Contributor
There was a problem hiding this comment.
There is a missing newline at the end of the file. Adding a trailing newline is a standard practice for text files to ensure compatibility with various tools and to follow POSIX standards.
Suggested change
| This issue has been observed on H100 with Qwen3-30B-A3B in colocate mode. | |
| This issue has been observed on H100 with Qwen3-30B-A3B in colocate mode. | |
- Update qa.md docs - Add disable_piecewise_cuda_graph argument - Adjust rollout worker cuda graph config
2bfaa2f to
d71b7cf
Compare
Remove models broken by transformers v5 tokenizer unification (DeepSeek-V3, step3, glm-4-9b-chat) and track them in a TOOL_CALL_KNOWN_FAILURES list with root cause comments. Add new passing models: Qwen3.5, Qwen3-Coder-Next, GLM-4.7-Flash, Kimi-K2.5, MiniMax-M2.5, Nemotron-3-Super. Clean up debug helpers. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
transformers >=5.0 changed apply_chat_template(tokenize=True) to return BatchEncoding instead of list[int]. Pass return_dict=False to all 6 call sites in mask_utils.py to ensure list[int] on both v4 and v5. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move Step-3.5-Flash from known failures into active tool-call test models, and clarify comments for remaining transformers v5 tokenizer/template incompatibilities. Made-with: Cursor
maocheng23
approved these changes
Apr 9, 2026
- Revert CI image back to radixark/miles:dev - Revert SGLANG_PR default back to sglang-miles - Revert SGLANG_BRANCH back to sglang-miles - Revert Megatron-Bridge back to merged-megatron-0.16.0rc0-miles
GuanxingLu
pushed a commit
to GuanxingLu/miles
that referenced
this pull request
Apr 21, 2026
DavidBellamy
added a commit
to LLM360/miles
that referenced
this pull request
Apr 21, 2026
…region clusters (#10) * Revert "[BUGFIX] [P2PRDMA] Add rollout post-processing after P2PRDMA weight updates" (radixark#882) * [Fix] fix ci (radixark#894) * Avoid threading for ray getting object (radixark#886) * Add explicit errors for unsupported Megatron profiles (radixark#887) * Add nvfp4 quantizer files (radixark#907) * Bump flash-linear-attention version to 0.4.2 (radixark#892) * [BUGFIX] Invoke "post_process_quantization" by default after weight updating (radixark#890) Co-authored-by: Yueming Yuan <yym022502@gmail.com> * Add heartbeat and id to session server (radixark#866) * fix: adding thin glm5 image to docker build + latest tag sync (radixark#871) * Add consistent hashing routing policy for rollout (radixark#891) Co-authored-by: Yueming Yuan <yueming@Mac.attlocal.net> * [example] add retool v2 example with multi-turn framework interfaces (radixark#654) Co-authored-by: GuanxingLu <gxlu02@gmail.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Expose rollout-batch-size, n-samples-per-prompt, global-batch-size as CLI args in swe-agent-v2 (radixark#954) Co-authored-by: Shi Dong <shi.dong@radixark.ai> * chore: remove obsolete swe-agent server.py and run-qwen3.sh (radixark#952) Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * Add weight staleness control for fully async rollout (radixark#958) * Fix/pause generation mode (radixark#924) Co-authored-by: Yueming Yuan <yym022502@gmail.com> * [v0.5.10][1] Bump sglang to v0.5.10 (radixark#898) * [v0.5.10][2] Fix apply_chat_template behavior for transformers >=5.0 (radixark#926) Co-authored-by: guapisolo <guapisolo@gmail.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * [v0.5.10][3] Fix processor return_tensors duplicate kwarg for transformers >=5.0 (radixark#927) Co-authored-by: guapisolo <guapisolo@gmail.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * [v0.5.10][4] Fix _no_split_modules set not subscriptable in transformers >=5.0 (radixark#931) * [v0.5.10][5] Disable piecewise cuda graph to avoid NVLS oom (radixark#935) * [v0.5.10][6][FSDP] fix outdated weight update logic in FSDP (radixark#948) Co-authored-by: guapisolo <guapisolo@gmail.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Co-authored-by: maocheng23 <35615230+maocheng23@users.noreply.github.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * [v0.5.10][7][FSDP] move FSDP to experimental and disable by default (radixark#961) * Add skiplist and more robust calculation on val (radixark#965) * [fix] tiny fix debug rollout only in weight version check (radixark#967) * feat: real cp support with relayout fix for qwen3.5 train/rollout mismatch (radixark#885) * [AMD] Upgrade to sglv0.5.10 (radixark#973) * switch model to actor (radixark#756) * [fix] support general logic to bypass fp32 downcast and fix qwen35 A_log dtype (radixark#975) Co-authored-by: yueming-yuan <yym022502@gmail.com> * fix: populate prefix_cache_info in OpenAI/session rollout path (radixark#960) * Remove prepare_harbor_tasks.py; use harbor-private adapters (radixark#982) * [fix] Skip flush_cache in in_place mode and add fully async example (radixark#974) Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * GLM47 full cmd for async and sync reasoning (radixark#986) Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> * fix: handle non-tool appended messages in TITO incremental tokenization (radixark#949) Co-authored-by: Yanbin Jiang <jybsuper@gmail.com> * [docker] Add sgl-model-gateway install and download .tar.gz assets (radixark#895) * [ci] fix hf rate limit error by caching tokenizer loading (radixark#1014) Co-authored-by: maocheng23 <35615230+maocheng23@users.noreply.github.com> * Use load_generate_function in legacy sglang_rollout path (radixark#1016) * Update CODEOWNERS to add new reviewers (radixark#1021) * Support moe lora for gpt-oss (radixark#798) Co-authored-by: Ethan (Yusheng) Su <yushengsu.thu@gmail.com> * [fix] restore expert_bias to fp32 before bridge weight export (radixark#811) * [chore] drop legacy transformers upgrade pin for glm47-flash and qwen35 (radixark#1018) * [fix] Enforce param dtype before wrap ddp (radixark#992) Co-authored-by: Zhichen Zeng <zczeng@uw.edu> * [upgrade] update Megatron-Bridge source and LoRA CI to megatron e2e tests and (radixark#1023) * [CI] Drop --use-miles-router from R3 tests and add r3 comparasion test between sgl & miles router (radixark#1015) * wandb: raise init_timeout, add retry wrapper, fix shared-mode init for cross-region clusters In online + shared mode, both `init_wandb_primary` and `init_wandb_secondary` make HTTPS round-trips to wandb cloud (login + run create/attach). On high-latency cross-region clusters (e.g. Abu Dhabi MBZUAI ↔ wandb-cloud US-West) with concurrent actor bursts, a single round-trip can exceed the wandb SDK's 90s default `init_timeout` — tearing down the whole run with a silent handshake abort. Observed on RL360 job 1564420, which forced `WANDB_MODE=offline` as a global default ever since (see https://github.com/LLM360/RL360/issues/87). The issue's original diagnosis assumed a local primary↔secondary socket handshake race. That's not how shared mode works — per wandb's own feature PR (wandb/wandb#6882), each writer spawns an independent wandb-core that talks to the cloud directly; aggregation is server-side by run_id. No local socket exists. The failure mode is pure network/latency, not a local readiness race. Changes ------- - Bump `init_timeout` to 300s for primary and secondary Settings. Configurable via `WANDB_INIT_TIMEOUT_SECS` env var for tuning. - Wrap both init paths in a bounded exponential-backoff retry (`_wandb_init_with_retry`) that re-attempts on wandb.errors.CommError and wandb.errors.UsageError. 3 attempts with 5→10→20s backoff by default, tunable via `WANDB_INIT_RETRY_ATTEMPTS` / `WANDB_INIT_RETRY_BACKOFF_SECS`. - Add `x_label` tagging per wandb distributed-training docs: primary gets `rank_<rank>_primary`, secondaries get `rank_<rank>_secondary`. Enables per-rank console-log filtering in the wandb UI. - Drop `reinit=True` from secondary init_kwargs. Shared mode natively supports concurrent writers on a single run; `reinit=True` triggered stale-state warnings on secondary actors without functional benefit. Followups this change enables ----------------------------- - `WANDB_MODE=offline` can be removed from scale.yaml's extra_env default once a pilot run confirms online mode boots cleanly. - The tmux-based `~/bin/wandb-sync-rl360.sh` workaround on David's M2 account becomes obsolete (no more offline-only default). - Near-realtime wandb dashboards replace the ~2-minute-lag offline sync; per-rank system metrics via x_label filtering. --------- Co-authored-by: JD <jaedon.guo@gmail.com> Co-authored-by: Ethan (Yusheng) Su <yushengsu.thu@gmail.com> Co-authored-by: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com> Co-authored-by: Ziang Li <ziangli@umich.edu> Co-authored-by: Zhichen Zeng <zczeng@uw.edu> Co-authored-by: JensenFire <xinji1@microsoft.com> Co-authored-by: Yueming Yuan <yym022502@gmail.com> Co-authored-by: maocheng23 <35615230+maocheng23@users.noreply.github.com> Co-authored-by: Douglas Yang <douglasyang88@gmail.com> Co-authored-by: Yueming Yuan <yueming@Mac.attlocal.net> Co-authored-by: Huapeng Zhou <73010314+PopSoda2002@users.noreply.github.com> Co-authored-by: GuanxingLu <gxlu02@gmail.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Co-authored-by: Shi-Dong <Shi-Dong@users.noreply.github.com> Co-authored-by: Shi Dong <shi.dong@radixark.ai> Co-authored-by: Jiajun Li <48857426+guapisolo@users.noreply.github.com> Co-authored-by: guapisolo <guapisolo@gmail.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Co-authored-by: Yuzhen Zhou <82826991+zyzshishui@users.noreply.github.com> Co-authored-by: Yanbin Jiang <jybsuper@gmail.com> Co-authored-by: Ying Sheng <sqy1415@gmail.com> Co-authored-by: Yisheng Gong <yishenggong9437@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Failed to bind NVLink SHARP (NVLS) Multicast memory--sglang-enforce-disable-flashinfer-allreduce-fusionto affected test scriptsRoot cause
flashinfer.comm.create_allreduce_fusion_workspace(backend="trtllm")in v0.5.10 uses NVLS multicast (vs IPC in v0.5.9). These handles persist through SGLang sleep and are not released for Megatron.DeepEP tests pass because they set
enable_flashinfer_allreduce_fusion=False.Test plan
test_qwen3_30B_A3B_r3.py(non-DeepEP) passestest_qwen3_30B_A3B.py(bf16, bridge) passes