Lucas Wilkinson ce75efeecb

[BugFix] FA2 MLA Accuracy Issue (#18807) ...

Signed-off-by: LucasWilkinson <lwilkinson@neuralmagic.com>

2025-05-28 08:59:39 +00:00

..

attention

[BugFix] FA2 MLA Accuracy Issue (#18807)

2025-05-28 08:59:39 +00:00

core

[Kernel] fp4 marlin kernel (#17687)

2025-05-10 19:58:49 -07:00

cpu

[Bugfix] Add half type support in reshape_and_cache_cpu_impl on x86 cpu platform (#18430)

2025-05-23 01:41:37 -07:00

cutlass_extensions

fix CUDA_check redefinition in #17918 (#18287)

2025-05-19 13:42:35 -07:00

mamba

feat(rocm-support): support mamba2 on rocm (#18565)

2025-05-27 00:07:53 -07:00

moe

[Build/CI] Fix CUDA 11.8 build (#17679)

2025-05-22 12:13:54 -07:00

prepare_inputs

[Misc][Easy] Annotate unused vars in the csrc files (#14798)

2025-03-15 12:40:09 +08:00

quantization

[Build/CI] Fix CUDA 11.8 build (#17679)

2025-05-22 12:13:54 -07:00

rocm

[ROCm][Kernel][V1] Enable AMD Radeon GPU Custom Paged Attention on v1 (#17004)

2025-05-21 08:35:00 -07:00

sparse/cutlass

fix CUDA_check redefinition in #17918 (#18287)

2025-05-19 13:42:35 -07:00

activation_kernels.cu

Modularize fused experts and integrate PPLX kernels (#15956)

2025-05-14 13:11:54 -07:00

cache_kernels.cu

Allocate kv_cache with stride order (#16605)

2025-04-25 22:03:31 -07:00

cache.h

[Attention] MLA with chunked prefill (#12639)

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

cuda_compat.h

[Kernel][ROCm][AMD] enable fused topk_softmax kernel for moe layer (#4927)

2024-06-02 14:13:26 -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

fix: spelling (#16466)

2025-04-11 23:24:22 -07:00

dispatch_utils.h

Modularize fused experts and integrate PPLX kernels (#15956)

2025-05-14 13:11:54 -07:00

layernorm_kernels.cu

[Kernel] Use fused rmsnorm for some models like qwen3 series (#17735)

2025-05-06 23:10:02 -07:00

layernorm_quant_kernels.cu

dynamic distpatch of fp8 kernels (#14245)

2025-03-11 10:54:56 -04:00

ops.h

Implements dual-chunk-flash-attn backend for dual chunk attention with sparse attention support (#11844)

2025-05-12 19:52:47 -07:00

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

[Kernel] Have rotary embeddings support tensors (#18046)

2025-05-14 15:43:55 -07:00

torch_bindings.cpp

feat(rocm-support): support mamba2 on rocm (#18565)

2025-05-27 00:07:53 -07:00

type_convert.cuh

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

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