Serendipity/CTI-Inference-Opt
1
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

perf: 降低合并阈值 0.85→0.80（继续探底）

Browse Source
This commit is contained in:
刘航宇
2026-06-14 12:09:28 +08:00
parent 2dcd74ba8f
commit e3590e6bda
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
代码/code/infer.py
+1 -1
View File
@@ -502,7 +502,7 @@ def load_model(ckpt_path, device='cuda:0'):
print("[INFO] Model converted to FP16 (embedding kept in FP32)") print("[INFO] Model converted to FP16 (embedding kept in FP32)")
# === 按 Expert 权重相似度合并冗余 expert === # === 按 Expert 权重相似度合并冗余 expert ===
_merge_experts(model, sim_threshold=0.85) _merge_experts(model, sim_threshold=0.80)
else: else:
print(f"[WARNING] Checkpoint {ckpt_path} not found, using random weights") print(f"[WARNING] Checkpoint {ckpt_path} not found, using random weights")