- 03 Dec, 2025 2 commits
- 02 Dec, 2025 1 commit
-
-
zhuwenwen authored
-
- 01 Dec, 2025 1 commit
-
-
zhuwenwen authored
[fix]解决deepseek模型cudagraph模式下精度异常问题 See merge request dcutoolkit/deeplearing/vllm!278
-
- 30 Nov, 2025 1 commit
-
-
王敏 authored
-
- 26 Nov, 2025 2 commits
- 25 Nov, 2025 1 commit
-
-
zhuwenwen authored
-
- 24 Nov, 2025 3 commits
- 21 Nov, 2025 1 commit
-
-
zhuwenwen authored
-
- 20 Nov, 2025 3 commits
- 18 Nov, 2025 2 commits
- 17 Nov, 2025 3 commits
- 14 Nov, 2025 2 commits
- 13 Nov, 2025 13 commits
-
-
zhuwenwen authored
[fix]解决moe_fused_gate编译错误 See merge request dcutoolkit/deeplearing/vllm!254
-
王敏 authored
-
zhuwenwen authored
set default_max_num_batched_tokens = 10240 update qwen3_moe of layernorm off lightop of moe_fused_gate
-
zhuwenwen authored
-
zhuwenwen authored
-
zhuwenwen authored
add VLLM_USE_PD_SPLIT to split prefill and decode replace triton_ of rms and act_and_mul
-
-
zhuwenwen authored
-
zhuwenwen authored
[fix]解决moe_fused_gate编译错误，去掉mla中mtp部分的修改 See merge request dcutoolkit/deeplearing/vllm!251
-
zhuwenwen authored
add deps of pd-pp
-
王敏 authored
-
王敏 authored
-
zhuwenwen authored
-
- 31 Oct, 2025 4 commits
- 30 Oct, 2025 1 commit
-
-
zhuwenwen authored
-
