Issue Station
Sign In
Tools hereToolbox

vllm - 💡(How to fix) Fix [Bug]: Gemma4 e2b and e4b models report 'Unsupported CPU attention configuration: head_dim=512 isa=1' error under vllm-cpu 0.19.0 [1 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#39065•Fetched 2026-04-08 02:52:40
View on GitHub
Comments
0
Participants
1
Timeline
3
Reactions
0
Author
chenqi123chenqi123
Participants
chenqi123chenqi123
Timeline (top)
closed ×1labeled ×1renamed ×1

Error Message

(EngineCore pid=148) Traceback (most recent call last):

Code Example

Your output of `python collect_env.py` here
RAW_BUFFERClick to expand / collapse

Your current environment

<details> <summary>The output of <code>python collect_env.py</code></summary> 
Your output of `python collect_env.py` here
</details>

🐛 Describe the bug

Starts up normally, but errors occur during inference service: (EngineCore pid=148) Traceback (most recent call last): (EngineCore pid=148) File "/opt/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap (EngineCore pid=148) self.run() (EngineCore pid=148) File "/opt/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 108, in run (EngineCore pid=148) self._target(*self._args, **self._kwargs) (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 1112, in run_engine_core (EngineCore pid=148) raise e (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 1101, in run_engine_core (EngineCore pid=148) engine_core.run_busy_loop() (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 1142, in run_busy_loop (EngineCore pid=148) self._process_engine_step() (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 1181, in _process_engine_step (EngineCore pid=148) outputs, model_executed = self.step_fn() (EngineCore pid=148) ^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/engine/core.py", line 392, in step (EngineCore pid=148) future = self.model_executor.execute_model(scheduler_output, non_block=True) (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/executor/uniproc_executor.py", line 114, in execute_model (EngineCore pid=148) output.result() (EngineCore pid=148) File "/opt/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/concurrent/futures/_base.py", line 449, in result (EngineCore pid=148) return self.__get_result() (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result (EngineCore pid=148) raise self._exception (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/executor/uniproc_executor.py", line 84, in collective_rpc (EngineCore pid=148) result = run_method(self.driver_worker, method, args, kwargs) (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/serial_utils.py", line 510, in run_method (EngineCore pid=148) return func(*args, **kwargs) (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/worker/worker_base.py", line 332, in execute_model (EngineCore pid=148) return self.worker.execute_model(scheduler_output) (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 124, in decorate_context (EngineCore pid=148) return func(*args, **kwargs) (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/worker/gpu_worker.py", line 803, in execute_model (EngineCore pid=148) output = self.model_runner.execute_model( (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 124, in decorate_context (EngineCore pid=148) return func(*args, **kwargs) (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/worker/gpu_model_runner.py", line 3970, in execute_model (EngineCore pid=148) self._build_attention_metadata( (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/worker/gpu_model_runner.py", line 2319, in _build_attention_metadata (EngineCore pid=148) _build_attn_group_metadata(kv_cache_gid, attn_gid, cm) (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/worker/gpu_model_runner.py", line 2270, in _build_attn_group_metadata (EngineCore pid=148) attn_metadata_i = builder.build( (EngineCore pid=148) ^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/v1/attention/backends/cpu_attn.py", line 173, in build (EngineCore pid=148) scheduler_metadata = ops.cpu_attn_get_scheduler_metadata( (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/vllm/_custom_ops.py", line 3113, in cpu_attn_get_scheduler_metadata (EngineCore pid=148) scheduler_metadata = torch.ops._C.get_scheduler_metadata( (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) File "/opt/venv/lib/python3.12/site-packages/torch/_ops.py", line 1209, in call (EngineCore pid=148) return self._op(*args, **kwargs) (EngineCore pid=148) ^^^^^^^^^^^^^^^^^^^^^^^^^ (EngineCore pid=148) RuntimeError: Unsupported CPU attention configuration: head_dim=512 isa=1

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

TL;DR

The most likely fix is to adjust the CPU attention configuration to a supported setup, potentially by modifying the head_dim or isa parameters.

Guidance

  • Verify the current CPU attention configuration and check if it matches the supported configurations.
  • Investigate the head_dim and isa parameters in the cpu_attn_get_scheduler_metadata function to determine the allowed values.
  • Consider updating the model or configuration to use a supported CPU attention setup.
  • Check the documentation and existing issues for any known workarounds or fixes for this specific error.

Example

No code snippet is provided as the issue is related to a specific configuration and the code is not explicitly mentioned.

Notes

The error message indicates an unsupported CPU attention configuration, but the exact solution depends on the specific requirements and constraints of the project. Further investigation and debugging may be necessary to determine the correct configuration.

Recommendation

Apply a workaround by adjusting the CPU attention configuration to a supported setup, as the error message suggests that the current configuration is not valid.

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…
#ISR setup#authentication setup#request error#file not found#serialization error

Still need to ship something?

×6

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

Back to top recommendations

TRENDING