Skip to content

GitLab

Unverified Commit e2fbfc95 authored by Divakar Verma's avatar Divakar Verma Committed by GitHub
Browse files

[CI][AMD] spec_decode:eagle skip FLASH_ATTN for deepseek on ROCm (#29827) 


Signed-off-by: default avatarDivakar Verma <divakar.verma@amd.com>
parent a690fb5b
Hide whitespace changes
Inline Side-by-side
Showing with 4 additions and 1 deletion
tests/v1/e2e/test_spec_decode.py
View file @ e2fbfc95
...@@ -414,7 +414,10 @@ def test_eagle_correctness( ...@@ -414,7 +414,10 @@ def test_eagle_correctness(
) )
if attn_backend == "FLASH_ATTN" and current_platform.is_rocm(): if attn_backend == "FLASH_ATTN" and current_platform.is_rocm():
m.setenv("VLLM_ROCM_USE_AITER", "1") if "deepseek" in model_setup[1].lower():
pytest.skip("FLASH_ATTN for deepseek not supported on ROCm platform")
else:
m.setenv("VLLM_ROCM_USE_AITER", "1")
method, model_name, spec_model_name, tp_size = model_setup method, model_name, spec_model_name, tp_size = model_setup
max_model_len = 2048 max_model_len = 2048
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment