dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix(critical): 3 bugs causing 0.0 score — OOM crash + thinking param + multimodal

Browse Source 
Bug 1 (FATAL): computility-run.yaml max-model-len 256000 → 131072,
gpu-memory-utilization 0.95 → 0.90, max-num-batched-tokens 4096 → 8192.
Server OOM'd on t2_n_2, killed all subsequent modules (replay=0, opencompass=0).

Bug 2 (functional): protocol.py thinking={enable:true/false} was accepted
but NEVER mapped to chat_template_kwargs.enable_thinking. Qwen3 template
never received the parameter → t1a, t1c, d07, d10 all FAIL.

Bug 3 (functional): chat_utils.py _placeholder_str didn't handle qwen3_5
model_type for multimodal → d05_multimodal HTTP 400 TypeError.

Expected: functional pass rate 0.41 → 0.90+, server stays alive for all
4 modules, total score 0.0 → 60000+ (matching reference sub 168).
This commit is contained in:
Claude
2026-08-07 07:05:40 +00:00
parent 539fe7745b
commit 16f0b30d2e
3 changed files with 18 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
qwen3_6_scripts/chat_utils.py
View File

@@ -172,7 +172,8 @@ class BaseMultiModalItemTracker(ABC, Generic[_T]):
return "<image>"
if model_type == "mllama":
return "<|image|>"
if model_type in ("qwen2_vl","qwen2_5_vl"):
if model_type in ("qwen2_vl", "qwen2_5_vl",
"qwen3_5", "qwen3_5_moe"):
return "<|vision_start|><|image_pad|><|vision_end|>"
if model_type == "molmo":
return ""
@@ -183,7 +184,8 @@ class BaseMultiModalItemTracker(ABC, Generic[_T]):
return "<|reserved_special_token_0|>"
raise TypeError(f"Unknown model type: {model_type}")
elif modality == "video":
if model_type in ("qwen2_vl","qwen2_5_vl"):
if model_type in ("qwen2_vl", "qwen2_5_vl",
"qwen3_5", "qwen3_5_moe"):
return "<|vision_start|><|video_pad|><|vision_end|>"
raise TypeError(f"Unknown model type: {model_type}")
else: