minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[CI] Gate Kimi-K3 acceptance length on the GSM8K average (#34145)

Browse Source
This commit is contained in:
Liangsheng Yin
2026-08-08 20:28:42 -07:00
committed by GitHub
parent 4792ab1e90
commit 1ebd6fab6c
1 changed files with 9 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/models_e2e/test_kimi_k3_b300.py
+9 -2
View File
@@ -40,8 +40,15 @@ class TestKimiK3B300LowLatency(GSM8KMixin, SpecDecodingMixin, CustomTestCase):
gsm8k_score_threshold = 0.95 gsm8k_score_threshold = 0.95
gsm8k_num_examples = 200 gsm8k_num_examples = 200
accept_length_thres = 6.6 # Gated on GSM8K rather than on test_bs_1_speed below: a 200-question
bs_1_speed_thres = 440 # average holds steady when a numerics change moves where the single
# greedy prompt hits EOS.
gsm8k_accept_length_thres = 4.5
# Both scale with how far that one greedy prompt runs, and speed is
# end-to-end, so launch and TTFT are amortized over the output -- it sits
# well below the steady decode rate the server logs. Coarse guards only.
accept_length_thres = 4.0
bs_1_speed_thres = 300
@classmethod @classmethod
def setUpClass(cls): def setUpClass(cls):