Alexander MatveevandGitHub 1210e4d95b [Bugfix] [B200] cutlass_mla - ensure kv_split == 1 for batch size > 1 (#25509) ...

Signed-off-by: Alexander Matveev <amatveev@redhat.com>

2025-09-23 16:57:55 -07:00

..

attention

[Bugfix] [B200] cutlass_mla - ensure kv_split == 1 for batch size > 1 (#25509)

2025-09-23 16:57:55 -07:00

core

…

cpu

Fix: Add explicit #include <omp.h> for OpenMP compatibility on certain toolchains (#24951)

2025-09-18 17:43:23 +08:00

cutlass_extensions

…

mamba/mamba_ssm

…

moe

Use macro guard CUDA functions for back compatibility in grouped_topk_kernel.cu (#25346)

2025-09-23 09:45:39 -07:00

quantization

[BugFix] Fix UB in per_token_group_quant.cu (#24913)

2025-09-23 09:14:22 -07:00

quickreduce

…

rocm

[ROCm] Add skinny gemm bias support for dtypes fp16,bf16,fp8 (#24988)

2025-09-23 14:31:45 -04:00

sparse/cutlass

…

activation_kernels.cu

…

cache_kernels.cu

…

cache.h

…

cub_helpers.h

Apply fixes for CUDA 13 (#24599)

2025-09-17 09:15:42 -04:00

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

…

launch_bounds_utils.h

[Compile] Fix Compile Warning for Ignoring MIN_BLOCK_PER_SM (#25193)

2025-09-19 16:23:19 -06:00

layernorm_kernels.cu

Apply fixes for CUDA 13 (#24599)

2025-09-17 09:15:42 -04:00

layernorm_quant_kernels.cu

Apply fixes for CUDA 13 (#24599)

2025-09-17 09:15:42 -04:00

ops.h

[Transform] Deterministic Hadacore Transforms (#24106)

2025-09-15 12:59:31 -06:00

permute_cols.cu

…

pos_encoding_kernels.cu

[Chore] Remove unused batched RoPE op & kernel (#24789)

2025-09-13 00:08:20 -07:00

sampler.cu

…

torch_bindings.cpp

Remove old cutlass mla (#23961)

2025-09-17 14:31:43 +00:00

type_convert.cuh

…