| deepseek v3/r1 and qwen support chunked prefill and prefix caching, kvcache input
| 1 年前 |
| 1. ds-r1 ep and add ut/st
2. mtp support 0.8.3
3. remove pynative judgement for the unit of eager mode and graph mode
| 11 个月前 |
| 1. ds-r1 ep and add ut/st
2. mtp support 0.8.3
3. remove pynative judgement for the unit of eager mode and graph mode
| 11 个月前 |
| 1. ds-r1 ep and add ut/st
2. mtp support 0.8.3
3. remove pynative judgement for the unit of eager mode and graph mode
| 11 个月前 |