Issue Station
Sign In
Tools hereToolbox

vllm - 💡(How to fix) Fix [Bug]: vllm start Qwen3.5-35B-A3B fail [2 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
vllm-project/vllm#36870•Fetched 2026-04-08 00:43:53
View on GitHub
Comments
2
Participants
2
Timeline
5
Reactions
0
Author
lcedawlcedaw
Participants
haosdenthaosdentlcedawlcedaw
Timeline (top)
commented ×2closed ×1labeled ×1subscribed ×1

Error Message

(EngineCore_DP0 pid=13824) Traceback (most recent call last):

RAW_BUFFERClick to expand / collapse

Your current environment

start: vllm serve /data/models/Qwen3.5-35B-A3B --host 0.0.0.0 --port 8000 --trust-remote-code --served-model-name qwen3.5-35b-a3b

🐛 Describe the bug

(EngineCore_DP0 pid=13824) Process EngineCore_DP0: (EngineCore_DP0 pid=13824) Traceback (most recent call last): (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap (EngineCore_DP0 pid=13824) self.run() (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/multiprocessing/process.py", line 108, in run (EngineCore_DP0 pid=13824) self._target(*self._args, **self._kwargs) (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 1010, in run_engine_core (EngineCore_DP0 pid=13824) raise e (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 997, in run_engine_core (EngineCore_DP0 pid=13824) engine_core = EngineCoreProc(*args, engine_index=dp_rank, **kwargs) (EngineCore_DP0 pid=13824) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 751, in init (EngineCore_DP0 pid=13824) super().init( (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 154, in init (EngineCore_DP0 pid=13824) num_gpu_blocks, num_cpu_blocks, kv_cache_config = self._initialize_kv_caches( (EngineCore_DP0 pid=13824) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 295, in _initialize_kv_caches (EngineCore_DP0 pid=13824) kv_cache_configs = get_kv_cache_configs( (EngineCore_DP0 pid=13824) ^^^^^^^^^^^^^^^^^^^^^ (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/core/kv_cache_utils.py", line 1505, in get_kv_cache_configs (EngineCore_DP0 pid=13824) global_kv_cache_groups = get_kv_cache_groups(vllm_config, merged_kv_cache_specs) (EngineCore_DP0 pid=13824) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/core/kv_cache_utils.py", line 1245, in get_kv_cache_groups (EngineCore_DP0 pid=13824) kv_cache_spec = unify_kv_cache_spec_page_size(kv_cache_spec) (EngineCore_DP0 pid=13824) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore_DP0 pid=13824) File "/usr/local/lib/python3.12/site-packages/vllm/v1/core/kv_cache_utils.py", line 939, in unify_kv_cache_spec_page_size (EngineCore_DP0 pid=13824) assert new_spec.page_size_bytes == max_page_size (EngineCore_DP0 pid=13824) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore_DP0 pid=13824) AssertionError

Before submitting a new issue...

  • Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

extent analysis

Fix Plan

The fix involves modifying the unify_kv_cache_spec_page_size function to handle cases where new_spec.page_size_bytes does not match max_page_size.

  • Modify the unify_kv_cache_spec_page_size function to remove the assertion and instead log a warning or error message when the page sizes do not match.
  • Update the get_kv_cache_configs function to handle cases where unify_kv_cache_spec_page_size returns an error.

Example code:

# In kv_cache_utils.py
def unify_kv_cache_spec_page_size(kv_cache_spec):
    new_spec = kv_cache_spec.copy()
    max_page_size = get_max_page_size()  # Assuming this function exists
    if new_spec.page_size_bytes != max_page_size:
        logger.warning(f"Page size mismatch: {new_spec.page_size_bytes} != {max_page_size}")
        # Optionally, return an error or set a default page size
        new_spec.page_size_bytes = max_page_size
    return new_spec

Verification

To verify the fix, run the vllm serve command with the same arguments as before and check for any error messages related to page size mismatches. The fix should prevent the AssertionError from occurring.

Extra Tips

  • Make sure to test the fix thoroughly to ensure it does not introduce any new issues.
  • Consider adding additional logging or monitoring to detect page size mismatches in the future.
  • If the page size mismatch is a common issue, consider updating the documentation to provide guidance on how to resolve it.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#ssr#installation#tensor shape#autograd error#model save/load#optimization#mixed precision#training loop

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING