Alexander Matveev 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

[Kernel] [Quantization] Add MXFP4 and bias support for marlin kernel (#22428)

2025-08-14 11:23:22 -07:00

cpu

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

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

cutlass_extensions

[Perf] Use upstream CUTLASS for SM90 Block FP8 kernel (#23280)

2025-09-11 15:43:14 -07:00

mamba/mamba_ssm

[V1][Mamba1] - FP32 SSM Kernel Support (#23506)

2025-09-01 20:53:00 -07:00

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

[Feature] add quick all reduce (#19744)

2025-06-26 20:54:24 -07:00

rocm

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

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

sparse/cutlass

[feat]: CUTLASS block scaled group gemm for SM100 (#19757)

2025-07-04 12:58:04 -06:00

activation_kernels.cu

[Kernel] Add cuda kernel for gpt_oss activation (#22951)

2025-08-17 05:03:24 +00:00

cache_kernels.cu

[Feature] Support Decode Context Parallel (DCP) for MLA (#23734)

2025-09-06 13:24:05 +08:00

cache.h

[Feature] Support Decode Context Parallel (DCP) for MLA (#23734)

2025-09-06 13:24:05 +08:00

cub_helpers.h

Apply fixes for CUDA 13 (#24599)

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

cuda_compat.h

[Bugfix][ROCm] Fix for warp_size uses on host (#21205)

2025-07-24 00:37:19 -07:00

cuda_utils_kernels.cu

[NVIDIA] Support nvfp4 quantization (#12784)

2025-02-12 19:51:51 -08:00

cuda_utils.h

[Attention] MLA with chunked prefill (#12639)

2025-02-21 15:30:12 -08:00

cuda_view.cu

[V1] Fully Transparent Implementation of CPU Offloading (#15354)

2025-03-31 20:22:34 +08:00

cumem_allocator.cpp

[core] improve error handling when wake up from sleep mode (#12981)

2025-02-10 09:38:57 +08:00

custom_all_reduce_test.cu

[Distributed] Add custom allreduce support for ROCM (#14125)

2025-03-31 22:49:12 -07:00

custom_all_reduce.cu

[Distributed] Add custom allreduce support for ROCM (#14125)

2025-03-31 22:49:12 -07:00

custom_all_reduce.cuh

[Kernels] Enable Torch Symmetric Memory All-Reduce By Default (#24111)

2025-09-11 09:45:31 -07:00

custom_quickreduce.cu

[Feature] add quick all reduce (#19744)

2025-06-26 20:54:24 -07:00

dispatch_utils.h

[Bugfix][Misc] Fix silu_and_mul_nvfp4_quant issue and extract common utils for nvfp4 kernel source files (#23727)

2025-09-04 14:25:45 -07:00

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

[Kernel] (2/N) Machete - Integrate into CompressedTensorsWNA16 and GPTQMarlin (#7701)

2024-09-23 13:46:26 -04:00

pos_encoding_kernels.cu

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

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

sampler.cu

[BUG] Fix #20484. Support empty sequence in cuda penalty kernel (#20491)

2025-07-05 19:38:02 -07:00

torch_bindings.cpp

Remove old cutlass mla (#23961)

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

type_convert.cuh

[torch.compile] Fuse RMSNorm with quant (#9138)

2024-11-08 21:20:08 +00:00