fzyzcjy 3b25dc127a [1/2] Speed up trtllm_mla attention backend (>10% e2e) (#10473)

2025-09-15 11:53:21 -07:00

..

activation.cu

…

cast.cu

…

concat_mla.cu

[1/2] Speed up trtllm_mla attention backend (>10% e2e) (#10473)

2025-09-15 11:53:21 -07:00

copy.cu

Support copying tensor from cpu to gpu without using copy engines (#10007)

2025-09-05 20:07:19 +08:00

fused_add_rms_norm_kernel.cu

…

pos_enc.cuh

Add PDL support for quant kernel and rope kernel (#9106)

2025-08-20 01:56:29 -07:00

rope.cu

Add PDL support for quant kernel and rope kernel (#9106)

2025-08-20 01:56:29 -07:00