[CI] Increase Qwen3.5 MTP GSM8K generation length (#49881)
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com> Co-authored-by: OpenAI Codex <codex@openai.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Jiangyun Zhu committed
Jul 28, 2026 at 18:04 UTC
25ace8fe5df07fc13f4aef5a89db391f326e60ee
1 file changed
+2
-1
tests/evals/gsm8k/configs/Qwen3.5-397B-A17B-NVFP4-DEP2-MTP.yaml
+2
-1
@@ -3,8 +3,9 @@ accuracy_threshold: 0.88
3
tolerance: 0.03
4
num_questions: 1319
5
num_fewshot: 5
6
+max_tokens: 12000
7
server_args: >-
7
- --max-model-len 4096
8
+ --max-model-len 16384
9
--data-parallel-size 2
10
--enable-expert-parallel
11
--max-num-seqs 384