[NPU][CI] Fail fast and speed up long-running qwen3.6 accuracy cases (#39813)

This commit is contained in:
pllimax
2026-09-16 20:20:37 +08:00
committed by GitHub
parent e7f7447333
commit 00a9a81b67
4 changed files with 12 additions and 3 deletions
@@ -91,6 +91,7 @@ class TestNPUQwen3_6_27B_1P_In64k_Out1k_Prefix90_gpqa(TestNpuAccuracyTestCaseBas
envs = QWEN3_6_27B_64K_PREFIX_ENVS
other_args = QWEN3_6_27B_64K_PREFIX_OTHER_ARGS
accuracy = 0.878
max_retries = 1
datasets = ["gpqa_diamond"]
few_shot_num = 0
eval_batch_size = 64
@@ -86,9 +86,10 @@ class TestNPUQwen3_6_27B_1P_In3k5_Out1k5_gpqa(TestNpuAccuracyTestCaseBase):
envs = QWEN3_6_27B_3K5_1K5_ENVS
other_args = QWEN3_6_27B_3K5_1K5_OTHER_ARGS
accuracy = 0.878
max_retries = 1
datasets = ["gpqa_diamond"]
few_shot_num = 0
eval_batch_size = 8
eval_batch_size = 16
generation_config = {
"max_tokens": 81920,
"temperature": 1.0,