Issue Station
Sign In
Tools hereToolbox

vllm - 💡(How to fix) Fix [Bug]: tests/lora/test_gptoss_tp.py::test_gpt_oss_lora_tp2[True-False] fails in eager mode

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…

Code Example

diff --git a/tests/lora/test_gptoss_tp.py b/tests/lora/test_gptoss_tp.py
index 68dd87233a..3fd1923c8a 100644
--- a/tests/lora/test_gptoss_tp.py
+++ b/tests/lora/test_gptoss_tp.py
@@ -129,9 +129,10 @@ def test_gpt_oss_lora_tp2(
             tensor_parallel_size=2,
             gpu_memory_utilization=0.8,
             fully_sharded_loras=fully_sharded_loras,
-            compilation_config=vllm.config.CompilationConfig(  # Avoid OOM
-                cudagraph_specialize_lora=False,
-            ),
+            # compilation_config=vllm.config.CompilationConfig(  # Avoid OOM
+            #    cudagraph_specialize_lora=False,
+            # ),
+            enforce_eager=True,
         )

         generate_and_test(llm, gptoss20b_lora_files, lora_id=1)
RAW_BUFFERClick to expand / collapse

Your current environment

env.txt

🐛 Describe the bug

tests/lora/test_gptoss_tp.py::test_gpt_oss_lora_tp2[True-False] fails if run in eager mode with an assertion about a non-contiguous tensor.

Modification to test_gptoss_tp.py

diff --git a/tests/lora/test_gptoss_tp.py b/tests/lora/test_gptoss_tp.py
index 68dd87233a..3fd1923c8a 100644
--- a/tests/lora/test_gptoss_tp.py
+++ b/tests/lora/test_gptoss_tp.py
@@ -129,9 +129,10 @@ def test_gpt_oss_lora_tp2(
             tensor_parallel_size=2,
             gpu_memory_utilization=0.8,
             fully_sharded_loras=fully_sharded_loras,
-            compilation_config=vllm.config.CompilationConfig(  # Avoid OOM
-                cudagraph_specialize_lora=False,
-            ),
+            # compilation_config=vllm.config.CompilationConfig(  # Avoid OOM
+            #    cudagraph_specialize_lora=False,
+            # ),
+            enforce_eager=True,
         )

         generate_and_test(llm, gptoss20b_lora_files, lora_id=1)

stack trace stack.txt

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 test failure in test_gptoss_tp.py can likely be resolved by adjusting the compilation configuration to accommodate eager mode.

Guidance

  • The introduction of enforce_eager=True in the test seems to be the cause of the failure, suggesting that the issue is related to the interaction between eager mode and the tensor parallelism.
  • Review the CompilationConfig settings, particularly cudagraph_specialize_lora, to ensure they are compatible with eager mode.
  • Consider re-enabling the commented-out CompilationConfig section with cudagraph_specialize_lora=False to see if it resolves the issue.
  • Verify that the test passes when run in non-eager mode to isolate the issue to eager mode specifically.

Example

No code snippet is provided as the necessary changes are implied to be within the existing configuration settings.

Notes

The exact solution may depend on the specifics of the vllm.config.CompilationConfig and its interaction with eager mode, which are not fully detailed in the issue.

Recommendation

Apply workaround: Re-enable the CompilationConfig with cudagraph_specialize_lora=False and adjust as necessary to accommodate eager mode, as this seems to be a direct approach to resolving the compatibility issue between eager mode and the current configuration.

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…
#authentication issue#prompt issue#agent setup#task chaining#parallel task

Still need to ship something?

×6

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

Back to top recommendations

TRENDING

vllm - 💡(How to fix) Fix [Bug]: tests/lora/test_gptoss_tp.py::test_gpt_oss_lora_tp2[True-False] fails in eager mode