| [Modify] Qwen3Omni: optimize memory with FlashAttention-2, activation offload and dynamic chunk loss | 4 个月前 |
| [Feature] Initialize Qwen3-Omni model code from transformers library | 3 个月前 |
| feat(torch): Qwen3-Omni support ulysses cp / fix(torch): repeat_kv and activation_offload bug | 3 个月前 |
| feat(torch): Qwen3-Omni support ulysses cp / fix(torch): repeat_kv and activation_offload bug | 3 个月前 |
| [Refactor]Refactor qwen3omni attention and use fusion op | 3 个月前 |