[release] v0.0.1

csrc/block_sparse_attn/src/flash_fwd_split_hdim128_bf16_sm80.cu

+// Copyright (c) 2023, Tri Dao.
+// Splitting the different head dimensions to different files to speed up compilation.
+// This file is auto-generated. See "generate_kernels.py"
+
+#include "flash_fwd_launch_template.h"
+
+template void run_mha_fwd_splitkv_dispatch<cutlass::bfloat16_t, 128>(Flash_fwd_params &params, cudaStream_t stream);