zhrrr 68c09efc37

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165) ...

Signed-off-by: zhuhaoran <zhuhaoran.zhr@alibaba-inc.com>

2025-11-11 12:00:31 -05:00

..

attention

[Bugfix][Kernel] fix merge attn states when both prefix and suffix are empty (#28181)

2025-11-06 17:52:13 +08:00

core

[small][batch invariance] Rename the env and internal flags to simplify usage (#26855)

2025-10-16 21:40:25 +00:00

cpu

[cpu][fix] Fix onednn_mm crash on consecutive matmuls with same M,K,N and different dtype (#27472)

2025-10-24 15:57:48 +00:00

cutlass_extensions

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

mamba/mamba_ssm

[V1] [Hybrid] Mamba1 Automatic Prefix Caching (#26377)

2025-11-02 04:16:23 -08:00

moe

[Perf][DeepSeek] Add sigmoid+bias fusion to fused_grouped_topk from TRTLLM (#28124)

2025-11-07 18:20:55 -08:00

quantization

Remove weight_scale.T special case for SM90 Block FP8 CUTLASS kernel (#28431)

2025-11-11 11:46:04 -05:00

quickreduce

[Bugfix][Rocm] fix qr error when different inp shape (#25892)

2025-10-13 10:04:21 -07:00

rocm

…

sparse/cutlass

…

activation_kernels.cu

…

cache_kernels.cu

…

cache.h

…

cub_helpers.h

…

cuda_compat.h

…

cuda_utils_kernels.cu

…

cuda_utils.h

…

cuda_view.cu

…

cumem_allocator.cpp

…

custom_all_reduce_test.cu

…

custom_all_reduce.cu

…

custom_all_reduce.cuh

…

custom_quickreduce.cu

…

dispatch_utils.h

…

fused_qknorm_rope_kernel.cu

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165)

2025-11-11 12:00:31 -05:00

launch_bounds_utils.h

…

layernorm_kernels.cu

[Chore] Remove unused PolyNorm layer (#27110)

2025-10-17 19:03:43 +00:00

layernorm_quant_kernels.cu

[torch.compile] Enable attention and allreduce fusion without custom ops enabled (#24604)

2025-10-17 08:10:23 -06:00

ops.h

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165)

2025-11-11 12:00:31 -05:00

permute_cols.cu

…

pos_encoding_kernels.cu

…

sampler.cu

[Deepseek v3.2] Remove extra logics in indexer (#26465)

2025-10-21 23:34:03 +00:00

torch_bindings.cpp

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165)

2025-11-11 12:00:31 -05:00

type_convert.cuh

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165)

2025-11-11 12:00:31 -05:00