Star
205
Fork
303
代码
介绍
代码
Issues
65
Pull Requests
96
流水线
Actions
项目讨论
Wiki
项目成员
89
分析
项目设置
Star
205
Fork
303
已合并
fix: use deterministic placeholder initialization for DeepSeek4 hash layers
#4898
丁子叉
创建于
17 天前
从
dingzicha1997/MindSpeed-LLM:
codex/hash-layer-init
合入到
Ascend/MindSpeed-LLM:
master
fix: use deterministic placeholder initialization for DeepSeek4 hash layers
#4898
已合并
丁子叉
创建于
17 天前
从
dingzicha1997/MindSpeed-LLM
codex/hash-layer-init
合入到
Ascend/MindSpeed-LLM
master
讨论
0
提交
0
检查
0
文件改动
0
克隆/下载
0
全选
2026-08-05
Revert "test: update DeepSeek4 ST baseline"
db508ee5
db508ee56f56cd063dba35d70bb265b80816268b
2026-08-04
test: update DeepSeek4 ST baseline
96017388
96017388094412033e7b5e013f710ce3851c6a37
2026-08-03
refactor: simplify router logits guard
c6916490
c691649032c8ce2ec6086bb309b3f40519e6b859
refactor: reduce lint suppressions in router
6dcb9817
6dcb981781ee5d0f986d348b3383985293e7f99f
fix: initialize hash router deterministically
4fb81b7c
4fb81b7ceb51d63262c0e927bd9d8de25432d81f
10
总计: 5
1