akhoroshev 78b6c4845a

Dynamically configure shared memory size for moe_align_block_size_kernel (#3376)

2024-03-14 18:18:07 -07:00

..

attention

[ROCM] Fix blockReduceSum to use correct warp counts for ROCm and CUDA (#3262)

2024-03-10 15:27:45 -07:00

moe

Add fused top-K softmax kernel for MoE (#2769)

2024-02-05 17:38:02 -08:00

punica

Add missing kernel for CodeLlama-34B on A/H100 (no tensor parallelism) when using Multi-LoRA. (#3350)

2024-03-13 12:18:25 -07:00

quantization

Integrate Marlin Kernels for Int4 GPTQ inference (#2497)

2024-03-01 12:47:51 -08:00

activation_kernels.cu

Add kernel for GeGLU with approximate GELU (#3337)

2024-03-12 22:06:17 -07:00

cache_kernels.cu

[Minor] Remove gather_cached_kv kernel (#3043)

2024-02-26 15:00:54 -08:00

cache.h

[Minor] Remove gather_cached_kv kernel (#3043)

2024-02-26 15:00:54 -08:00

cuda_compat.h

[ROCM] Fix blockReduceSum to use correct warp counts for ROCm and CUDA (#3262)

2024-03-10 15:27:45 -07:00

cuda_utils_kernels.cu

[ROCm] add support to ROCm 6.0 and MI300 (#2274)

2024-01-26 12:41:10 -08:00

cuda_utils.h

[ROCm] add support to ROCm 6.0 and MI300 (#2274)

2024-01-26 12:41:10 -08:00

custom_all_reduce_test.cu

Implement custom all reduce kernels (#2192)

2024-01-27 12:46:35 -08:00

custom_all_reduce.cu

Implement custom all reduce kernels (#2192)

2024-01-27 12:46:35 -08:00

custom_all_reduce.cuh

No repeated IPC open (#2642)

2024-01-29 10:46:29 -08:00

dispatch_utils.h

DeepseekMoE support with Fused MoE kernel (#2453)

2024-01-29 21:19:48 -08:00

layernorm_kernels.cu

[FIX] Support non-zero CUDA devices in custom kernels (#1959)

2024-01-02 19:09:59 -08:00

moe_align_block_size_kernels.cu

Dynamically configure shared memory size for moe_align_block_size_kernel (#3376)

2024-03-14 18:18:07 -07:00

ops.h

Add batched RoPE kernel (#3095)

2024-03-13 13:45:26 -07:00

pos_encoding_kernels.cu

Add batched RoPE kernel (#3095)

2024-03-13 13:45:26 -07:00

pybind.cpp

Add batched RoPE kernel (#3095)

2024-03-13 13:45:26 -07:00

reduction_utils.cuh

[ROCm] Fix warp and lane calculation in blockReduceSum (#3321)

2024-03-11 13:14:07 -07:00