* LLM: optimize llama natvie sdp for split qkv tensor. * fix block real size. * fix comment. * fix style. * refactor. |
||
|---|---|---|
| .. | ||
| llm | ||
* LLM: optimize llama natvie sdp for split qkv tensor. * fix block real size. * fix comment. * fix style. * refactor. |
||
|---|---|---|
| .. | ||
| llm | ||