Skip to content

[Fix] Fixed Qwen3.5 dense model load weight issue (#232).#236

Merged
qichu-yun merged 1 commit intozejunchen-zejun:Qwen3.5_v0.5.9from
IzacharyI:Qwen3.5_v0.5.9
Apr 3, 2026
Merged

[Fix] Fixed Qwen3.5 dense model load weight issue (#232).#236
qichu-yun merged 1 commit intozejunchen-zejun:Qwen3.5_v0.5.9from
IzacharyI:Qwen3.5_v0.5.9

Conversation

@IzacharyI
Copy link
Copy Markdown

Motivation

Fixed dense model load weight issue (#232) .

Modifications

Cherry-pick PR sgl-project#21019 and Add BF16 qkv z b a fusion and PTPC quant config

Accuracy Tests

397B-A17B BF16:
image
397B-A17B PTPC FP8:
image
27B BF16

  • TP1:
image
  • TP 2:
image

Benchmarking and Profiling

Checklist

- Cherry-pick PR sgl-project#21019: Fuse GDN split/reshape/cat ops with FP8/BF16 quant support
- Add BF16 qkv z b a fusion and PTPC quant config
@qichu-yun qichu-yun merged commit 9c0cea4 into zejunchen-zejun:Qwen3.5_v0.5.9 Apr 3, 2026
qichu-yun added a commit that referenced this pull request Apr 3, 2026
qichu-yun added a commit that referenced this pull request Apr 3, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants