 Yongfei Xu
|
ebe3ab29e4
|
[DeepSeek V4] CP decode opt: slice repeat attention weights to local TP partition (#27657)
|
2026-07-23 14:06:25 -07:00 |
|
|
Yongfei Xu
|
98eb84497d
|
[PP] Skip PP output communication for pure chunked prefill batches (#26148)
|
2026-05-26 21:59:18 +08:00 |
|
|
Yongfei Xu
|
17031120b8
|
[DeepSeek v3.2][Bugfix] get_index_k_scale_buffer support cp (#18280)
|
2026-03-17 09:54:54 -07:00 |
|
|
Yongfei Xu
|
677f3c49da
|
[DeepSeek V3.2] [Bugfix] slice indexer and padding fa3 when can not run cuda graph (#17076)
|
2026-02-03 01:32:20 +08:00 |
|
|
Yongfei Xu
|
d2105d4abd
|
[DeepSeek v3.2] Opt MTP decode cuda batch sizes and nsa implementation (#16961)
|
2026-01-19 11:54:11 +08:00 |
|
|
Yongfei Xu
|
82a1b645ba
|
[DeepSeek V3.1/V3.2] Optimize fused moe configs for H20 & H20-3E based on swapab (#17133)
|
2026-01-17 00:10:52 +08:00 |
|
|
Yongfei Xuandybyang
|
05dfef92a1
|
[DeepSeek 3.2] Support and optimize pipeline parallelis when context pipeline enabled (#16380)
Co-authored-by: ybyang <ybyang7@iflytek.com>
|
2026-01-09 11:01:49 +08:00 |
|
|
Yongfei Xu
|
0d244116d2
|
[DeepSeek v3.2] opt Context Parallelism: support fused moe, multi batch and fp8 kvcache (#13959)
|
2026-01-02 23:49:14 +08:00 |
|
|
Yongfei Xu
|
d91b16eb16
|
Opt tp: tp attn support tp reduce scattered input (#10568)
|
2025-11-15 02:08:12 -08:00 |
|
|
Yongfei Xuandybyang
|
d2b8c4123e
|
Opt fused triton moe: add tma for down proj kernel (#10567)
Co-authored-by: ybyang <10629930+whybeyoung@users.noreply.github.com>
|
2025-10-28 14:26:17 +08:00 |
|
|
Yongfei Xu
|
4793ec7d1a
|
Opt MHA chunked prefix: merge prefix and extend kv cache to run mha once (#10953)
|
2025-10-23 20:58:10 -07:00 |
|