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

fix: adjusting vlm accuracy thresholds (#16593)

Browse Source
This commit is contained in:
Douglas Yang
2026-01-06 19:41:05 -08:00
committed by GitHub
parent 53846746bf
commit 951d16c890
1 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/eval/test_vlms_mmmu_eval.py
+6 -6
View File
@@ -34,24 +34,24 @@ MODEL_THRESHOLDS = {
ModelLaunchSettings("google/gemma-3n-E4B-it"): ModelEvalMetrics(0.270, 17.7), ModelLaunchSettings("google/gemma-3n-E4B-it"): ModelEvalMetrics(0.270, 17.7),
ModelLaunchSettings("mistral-community/pixtral-12b"): ModelEvalMetrics(0.360, 16.6), ModelLaunchSettings("mistral-community/pixtral-12b"): ModelEvalMetrics(0.360, 16.6),
ModelLaunchSettings("moonshotai/Kimi-VL-A3B-Instruct"): ModelEvalMetrics( ModelLaunchSettings("moonshotai/Kimi-VL-A3B-Instruct"): ModelEvalMetrics(
0.330, 22.3 0.330, 23.5
), ),
ModelLaunchSettings("openbmb/MiniCPM-o-2_6"): ModelEvalMetrics(0.330, 29.3), ModelLaunchSettings("openbmb/MiniCPM-o-2_6"): ModelEvalMetrics(0.330, 29.5),
ModelLaunchSettings("openbmb/MiniCPM-v-2_6"): ModelEvalMetrics(0.259, 36.3), ModelLaunchSettings("openbmb/MiniCPM-v-2_6"): ModelEvalMetrics(0.259, 36.3),
ModelLaunchSettings("OpenGVLab/InternVL2_5-2B"): ModelEvalMetrics(0.300, 17.0), ModelLaunchSettings("OpenGVLab/InternVL2_5-2B"): ModelEvalMetrics(0.300, 18.0),
ModelLaunchSettings("Qwen/Qwen2-VL-7B-Instruct"): ModelEvalMetrics(0.310, 83.3), ModelLaunchSettings("Qwen/Qwen2-VL-7B-Instruct"): ModelEvalMetrics(0.310, 83.3),
ModelLaunchSettings("Qwen/Qwen2.5-VL-7B-Instruct"): ModelEvalMetrics(0.340, 31.9), ModelLaunchSettings("Qwen/Qwen2.5-VL-7B-Instruct"): ModelEvalMetrics(0.330, 31.9),
ModelLaunchSettings( ModelLaunchSettings(
"Qwen/Qwen3-VL-30B-A3B-Instruct", extra_args=["--tp=2"] "Qwen/Qwen3-VL-30B-A3B-Instruct", extra_args=["--tp=2"]
): ModelEvalMetrics(0.29, 37.0), ): ModelEvalMetrics(0.29, 37.0),
ModelLaunchSettings( ModelLaunchSettings(
"unsloth/Mistral-Small-3.1-24B-Instruct-2503" "unsloth/Mistral-Small-3.1-24B-Instruct-2503"
): ModelEvalMetrics(0.310, 16.7), ): ModelEvalMetrics(0.30, 16.7),
ModelLaunchSettings("XiaomiMiMo/MiMo-VL-7B-RL"): ModelEvalMetrics(0.28, 32.0), ModelLaunchSettings("XiaomiMiMo/MiMo-VL-7B-RL"): ModelEvalMetrics(0.28, 32.0),
ModelLaunchSettings("zai-org/GLM-4.1V-9B-Thinking"): ModelEvalMetrics(0.280, 30.4), ModelLaunchSettings("zai-org/GLM-4.1V-9B-Thinking"): ModelEvalMetrics(0.280, 30.4),
ModelLaunchSettings( ModelLaunchSettings(
"zai-org/GLM-4.5V-FP8", extra_args=["--tp=2"] "zai-org/GLM-4.5V-FP8", extra_args=["--tp=2"]
): ModelEvalMetrics(0.26, 32.0), ): ModelEvalMetrics(0.26, 34.0),
} }