Mojo module
bmm
comptime valuesβ
elementwise_epilogue_typeβ
comptime elementwise_epilogue_type = def[c_type: DType, width: Int, rank: Int, *, alignment: Int = 1](IndexList[rank], SIMD[c_type, width]) capturing -> None
loggerβ
comptime logger = Logger(stdout, prefix=String(""), source_location=False)
Functionsβ
- β
batched_matmul: TileTensor primary implementation ofbatched_matmul. - β
batched_matmul_dynamic_scaled_fp8: - β
batched_matmul_dynamic_scaled_fp8_naive: - β
batched_matmul_kernel_gpu: - β
batched_matmul_shape: Compute the output shape of abatch_matmuloperation, and assert the inputs are compatible. - β
bmm_sm100_blockwise_scaled_fp8: - β
naive_batched_matmul_kernel:
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!