kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

fix: [https://nvbugs/5355219] Fix bug of Qwen3 235B CI on dgx_gb200 (#5602)

Browse Source 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
This commit is contained in:
bhsueh_NV 2025-07-02 10:07:01 +08:00 committed by GitHub
parent 682b164b9b
commit d5606b062a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tests/integration/defs/accuracy/test_llm_api_pytorch.py
View File

@ -1457,13 +1457,15 @@ class TestQwen3_235B_A22B(LlmapiAccuracyTestHarness):
pytorch_config = dict(disable_overlap_scheduler=not overlap_scheduler,
use_cuda_graph=cuda_graph)
kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.6)
llm = LLM(
f"{llm_models_root()}/Qwen3/saved_models_Qwen3-235B-A22B_nvfp4_hf",
tensor_parallel_size=tp_size,
pipeline_parallel_size=pp_size,
moe_expert_parallel_size=ep_size,
**pytorch_config,
enable_attention_dp=attention_dp)
enable_attention_dp=attention_dp,
kv_cache_config=kv_cache_config)
with llm:
task = MMLU(self.MODEL_NAME)
task.evaluate(llm)