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

[NPU] bugfix with Kimi-k2 and bge-reranker-v2 model (#17478)

Browse Source 
Co-authored-by: amote-i <49533125+amote-i@users.noreply.github.com>
Co-authored-by: cy <chenyang08056032@163.com>
This commit is contained in:
chenxu214
2026-01-22 22:02:05 +08:00
committed by GitHub
co-authored by amote-i cy
parent a4dc432587
commit 5d299c25c0
5 changed files with 130 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py
+1 -1
View File
@@ -1648,7 +1648,7 @@ class NPUCompressedTensorsW4A16Int4DynamicMoEMethod(CompressedTensorsMoEMethod):
self.num_experts = num_experts
if (
extra_weight_attrs.get(
"intermediate_size_full", intermediate_size_per_partition
"moe_intermediate_size", intermediate_size_per_partition
)
// intermediate_size_per_partition
> 1