minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Relax flaky test thresholds for MLA DeepSeek V3 and AutoRound (#20068)

Browse Source 
Co-authored-by: Alison Shao <alisonshao@Mac.attlocal.net>
This commit is contained in:
Alison Shao
2026-03-06 17:26:28 -08:00
committed by GitHub
co-authored by Alison Shao
parent a0d085c16d
commit 50bbdcf8e9
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/mla/test_mla_deepseek_v3.py
+1 -1
View File
@@ -144,7 +144,7 @@ class TestMLADeepseekV3Fa3Fp8Kvcache(CustomTestCase):
metrics = run_eval_few_shot_gsm8k(args) metrics = run_eval_few_shot_gsm8k(args)
print(metrics) print(metrics)
self.assertGreater(metrics["accuracy"], 0.62) self.assertGreater(metrics["accuracy"], 0.60)
class TestDeepseekV3MTP(CustomTestCase): class TestDeepseekV3MTP(CustomTestCase):
test/registered/quant/test_autoround.py
+1 -1
View File
@@ -55,7 +55,7 @@ class TestAutoRound(CustomTestCase):
if "Llama" in model: if "Llama" in model:
self.assertGreaterEqual(metrics["score"], 0.6) self.assertGreaterEqual(metrics["score"], 0.6)
else: else:
self.assertGreaterEqual(metrics["score"], 0.26) self.assertGreaterEqual(metrics["score"], 0.25)
finally: finally:
kill_process_tree(process.pid) kill_process_tree(process.pid)
print(f"[INFO] Server for {model} stopped.") print(f"[INFO] Server for {model} stopped.")