Star204
300
代码
代码
Issues59
Pull Requests88
流水线
项目讨论
Wiki
分析
Star204
300
已合并
refactor: remove deepseek4 redundant Triton paths and align CSA/MHC CLI #4871
丁子叉创建于 21 天前
refactor: remove deepseek4 redundant Triton paths and align CSA/MHC CLI #4871
已合并
丁子叉创建于 21 天前
从dingzicha1997/MindSpeed-LLMremove-triton-mhc合入到Ascend/MindSpeed-LLMmaster
2026-07-30
fix: initialize MindSpeed before Megatron imports
Ddingzicha1997提交于 19 天前
d3179b22
fix: allow DeepSeek V4 position embedding with MTP
Ddingzicha1997提交于 19 天前
a991670e
merge: sync master into remove-triton-mhc
Ddingzicha1997提交于 19 天前
295a4ca8
2026-07-29
style: satisfy DeepSeek V4 pre-commit checks
Ddingzicha1997提交于 20 天前
f5388248
fix: remove legacy rotary flags from DeepSeek V4 scripts
Ddingzicha1997提交于 20 天前
010cce73
fix: synchronize fused lightning indexer flags
Ddingzicha1997提交于 20 天前
53d7206e
refactor: relocate DeepSeek V4 MHC operators
Ddingzicha1997提交于 20 天前
49f55ad3
2026-07-28
refactor: align DeepSeek V4 CSA naming and arguments
Ddingzicha1997提交于 21 天前
ebdac8cd
refactor: remove Triton CSA attention path
Ddingzicha1997提交于 21 天前
712b7c1d
refactor: consolidate DeepSeek V4 attention code
Ddingzicha1997提交于 21 天前
4befccb2
总计: 12