biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[ROCm][CI] Prep Tests For Change To ROCM_ATTN As New Default Backend On ROCm (#36025)

Browse Source 
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
This commit is contained in:
Micah Williamson
2026-03-09 13:27:55 -05:00
committed by GitHub
parent 3fd03f1ec2
commit 4ff9b045fe
10 changed files with 32 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/v1/e2e/test_spec_decode.py
View File

@@ -732,11 +732,13 @@ def test_mtp_correctness(
method, model_name, tp_size = model_setup
_skip_if_insufficient_gpus_for_tp(tp_size)
attn_backend = "TRITON_ATTN" if current_platform.is_rocm() else "auto"
ref_llm = LLM(
model=model_name,
max_model_len=2048,
tensor_parallel_size=tp_size,
trust_remote_code=True,
attention_backend=attn_backend,
)
ref_outputs = ref_llm.chat(test_prompts, sampling_config)
evaluate_llm_for_gsm8k(
@@ -756,6 +758,7 @@ def test_mtp_correctness(
"max_model_len": 2048,
},
max_model_len=2048,
attention_backend=attn_backend,
)
evaluate_llm_for_gsm8k(
spec_llm, expected_accuracy_threshold=expected_accuracy_threshold