Jinzhen Lin 750f4cabfa

[Kernel] optimize moe_align_block_size for cuda graph and large num_experts (e.g. DeepSeek-V3) (#12222) ...

Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
Co-authored-by: Michael Goin <mgoin@redhat.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>

2025-01-20 16:42:16 -08:00

..

attention

[MISC] Replace c10::optional with std::optional (#11730)

2025-01-05 10:20:34 +09:00

core

Move linting to pre-commit (#11975)

2025-01-20 14:58:01 +08:00

cpu

Move linting to pre-commit (#11975)

2025-01-20 14:58:01 +08:00

cutlass_extensions

Move linting to pre-commit (#11975)

2025-01-20 14:58:01 +08:00

mamba

[MISC] Replace c10::optional with std::optional (#11730)

2025-01-05 10:20:34 +09:00

moe

[Kernel] optimize moe_align_block_size for cuda graph and large num_experts (e.g. DeepSeek-V3) (#12222)

2025-01-20 16:42:16 -08:00

prepare_inputs

[Misc] Add multipstep chunked-prefill support for FlashInfer (#10467)

2025-01-15 12:47:49 +08:00

quantization

[MISC] Replace c10::optional with std::optional (#11730)

2025-01-05 10:20:34 +09:00

rocm

[MISC] Replace c10::optional with std::optional (#11730)

2025-01-05 10:20:34 +09:00

sparse/cutlass

[MISC] Replace c10::optional with std::optional (#11730)

2025-01-05 10:20:34 +09:00

activation_kernels.cu

[Kernel] Support MulAndSilu (#11624)

2025-01-15 02:29:53 +00:00

cache_kernels.cu

[V1] Further reduce CPU overheads in flash-attn (#10989)

2024-12-09 12:38:46 -08:00

cache.h

Add fp8 support to reshape_and_cache_flash (#6667)

2024-07-24 18:36:52 +00:00

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

[Kernel][Misc] Use TORCH_LIBRARY instead of PYBIND11_MODULE for custom ops (#5047)

2024-06-09 16:23:30 -04:00

cuda_utils.h

[Kernel] (1/N) Machete - Hopper Optimized Mixed Precision Linear Kernel (#7174)

2024-08-20 07:09:33 -06:00

custom_all_reduce_test.cu

[Core][Distributed] Refactor ipc buffer init in CustomAllreduce (#10030)

2024-11-06 23:50:47 -08:00

custom_all_reduce.cu

[Core][Distributed] Refactor ipc buffer init in CustomAllreduce (#10030)

2024-11-06 23:50:47 -08:00

custom_all_reduce.cuh

[Core][Distributed] Refactor ipc buffer init in CustomAllreduce (#10030)

2024-11-06 23:50:47 -08:00

dispatch_utils.h

[torch.compile] Dynamic fp8 + rms_norm fusion (#10906)

2024-12-13 03:19:23 +00:00

layernorm_kernels.cu

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

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

layernorm_quant_kernels.cu

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

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

ops.h

[Kernel] Support MulAndSilu (#11624)

2025-01-15 02:29:53 +00:00

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][Misc] Use TORCH_LIBRARY instead of PYBIND11_MODULE for custom ops (#5047)

2024-06-09 16:23:30 -04:00

torch_bindings.cpp

[Kernel] Support MulAndSilu (#11624)

2025-01-15 02:29:53 +00:00

type_convert.cuh

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

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