…

..

awq_kernel.cu

…

bmm_fp8.cu

…

fp8_blockwise_gemm_kernel.cu

…

fp8_gemm_kernel.cu

…

int8_gemm_kernel.cu

…

nvfp4_quant_entry.cu

…

nvfp4_quant_kernels.cu

…

nvfp4_scaled_mm_entry.cu

…

nvfp4_scaled_mm_kernels.cu

…

per_tensor_quant_fp8.cu

…

per_token_group_quant_8bit.cu

…

per_token_quant_fp8.cu

Speed up per token and per tensor quant by 15% (#4639)

2025-03-22 00:37:57 -07:00