AUTHORS
LICENSE
MANIFEST.in
README.md
pyproject.toml
setup.py
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/fag_general_host.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/flash_api.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_fp16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_fp16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/varlen_bwd_dispatch_bf16.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu/autogen/varlen_bwd_dispatch_fp16.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/flash_api.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_fp16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_fp16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/flash_api.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_fp16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_fp16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_3/fai_host_api.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_3/flash_api.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_fp16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_4/fai_host_api.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_4/flash_api.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_bf16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_bf16_tnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_fp16_bsnd.cpp
/data0/liaojy/workspace/FA/flash-attention-npu/csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_fp16_tnd.cpp
csrc/ascend910/flash_attn_npu/CombineScale.hpp
csrc/ascend910/flash_attn_npu/fa_block.h
csrc/ascend910/flash_attn_npu/fag_block.h
csrc/ascend910/flash_attn_npu/fag_epilogue_op.hpp
csrc/ascend910/flash_attn_npu/fag_epilogue_post.hpp
csrc/ascend910/flash_attn_npu/fag_epilogue_pre.hpp
csrc/ascend910/flash_attn_npu/fag_epilogue_sfmg.hpp
csrc/ascend910/flash_attn_npu/fag_general_dispatch.hpp
csrc/ascend910/flash_attn_npu/fag_general_dispatch_impl.hpp
csrc/ascend910/flash_attn_npu/fag_general_host.cpp
csrc/ascend910/flash_attn_npu/fag_general_host.hpp
csrc/ascend910/flash_attn_npu/fag_general_launch.hpp
csrc/ascend910/flash_attn_npu/fag_mmad_cube1.hpp
csrc/ascend910/flash_attn_npu/fag_mmad_cube2.hpp
csrc/ascend910/flash_attn_npu/fag_mmad_cube3.hpp
csrc/ascend910/flash_attn_npu/fag_sfmg.h
csrc/ascend910/flash_attn_npu/fag_tiling.cpp
csrc/ascend910/flash_attn_npu/flash_api.cpp
csrc/ascend910/flash_attn_npu/fwd_dispatch.hpp
csrc/ascend910/flash_attn_npu/fwd_dispatch_impl.hpp
csrc/ascend910/flash_attn_npu/init_outputs.hpp
csrc/ascend910/flash_attn_npu/kernel_common.hpp
csrc/ascend910/flash_attn_npu/kernel_common_fag.hpp
csrc/ascend910/flash_attn_npu/mha_fwd_kvcache.cpp
csrc/ascend910/flash_attn_npu/mha_varlen_bwd.cpp
csrc/ascend910/flash_attn_npu/online_softmax.hpp
csrc/ascend910/flash_attn_npu/pv_matmul.hpp
csrc/ascend910/flash_attn_npu/qk_matmul.hpp
csrc/ascend910/flash_attn_npu/rescale_o.hpp
csrc/ascend910/flash_attn_npu/softmax_tiling.cpp
csrc/ascend910/flash_attn_npu/tilingdata.h
csrc/ascend910/flash_attn_npu/varlen_bwd_dispatch.hpp
csrc/ascend910/flash_attn_npu/varlen_bwd_dispatch_impl.hpp
csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_bf16_bsnd.cpp
csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_bf16_tnd.cpp
csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_fp16_bsnd.cpp
csrc/ascend910/flash_attn_npu/autogen/fag_general_dispatch_fp16_tnd.cpp
csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_bf16_bsnd.cpp
csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_bf16_tnd.cpp
csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_fp16_bsnd.cpp
csrc/ascend910/flash_attn_npu/autogen/fwd_dispatch_fp16_tnd.cpp
csrc/ascend910/flash_attn_npu/autogen/varlen_bwd_dispatch_bf16.cpp
csrc/ascend910/flash_attn_npu/autogen/varlen_bwd_dispatch_fp16.cpp
csrc/ascend910/flash_attn_npu/fag_common/common_header.h
csrc/ascend910/flash_attn_npu/fag_common/cube_addr.h
csrc/ascend910/flash_attn_npu/fag_common/vector_addr.h
csrc/ascend910/flash_attn_npu_3/CombineScale.hpp
csrc/ascend910/flash_attn_npu_3/bwd_dispatch.hpp
csrc/ascend910/flash_attn_npu_3/bwd_dispatch_common.hpp
csrc/ascend910/flash_attn_npu_3/fa_block.h
csrc/ascend910/flash_attn_npu_3/fa_metadata.aicpu
csrc/ascend910/flash_attn_npu_3/fa_metadata_args.h
csrc/ascend910/flash_attn_npu_3/fa_split.h
csrc/ascend910/flash_attn_npu_3/fag_kernel.cpp
csrc/ascend910/flash_attn_npu_3/fag_tiling.cpp
csrc/ascend910/flash_attn_npu_3/fag_tiling.h
csrc/ascend910/flash_attn_npu_3/flash_api.cpp
csrc/ascend910/flash_attn_npu_3/fwd_dispatch.hpp
csrc/ascend910/flash_attn_npu_3/fwd_dispatch_impl.hpp
csrc/ascend910/flash_attn_npu_3/init_outputs.hpp
csrc/ascend910/flash_attn_npu_3/kernel_common.hpp
csrc/ascend910/flash_attn_npu_3/mha_fwd_kvcache.cpp
csrc/ascend910/flash_attn_npu_3/online_softmax.hpp
csrc/ascend910/flash_attn_npu_3/pv_matmul.hpp
csrc/ascend910/flash_attn_npu_3/qk_matmul.hpp
csrc/ascend910/flash_attn_npu_3/rescale_o.hpp
csrc/ascend910/flash_attn_npu_3/tilingdata.h
csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_bf16_bsnd.cpp
csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_bf16_tnd.cpp
csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_fp16_bsnd.cpp
csrc/ascend910/flash_attn_npu_3/autogen/bwd_dispatch_fp16_tnd.cpp
csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_bf16_bsnd.cpp
csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_bf16_tnd.cpp
csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_fp16_bsnd.cpp
csrc/ascend910/flash_attn_npu_3/autogen/fwd_dispatch_fp16_tnd.cpp
csrc/ascend910/flash_attn_npu_3/fag_common/epilogue_fag_deterministic_add.hpp
csrc/ascend910/flash_attn_npu_3/fag_common/mmad_fag_dqkv.hpp
csrc/ascend910/flash_attn_npu_3/fag_common/mmad_fag_sdp.hpp
csrc/ascend910/flash_attn_npu_4/CombineScale.hpp
csrc/ascend910/flash_attn_npu_4/bwd_dispatch.hpp
csrc/ascend910/flash_attn_npu_4/bwd_dispatch_common.hpp
csrc/ascend910/flash_attn_npu_4/fa_block.h
csrc/ascend910/flash_attn_npu_4/fa_split.h
csrc/ascend910/flash_attn_npu_4/fag_kernel.cpp
csrc/ascend910/flash_attn_npu_4/fag_tiling.cpp
csrc/ascend910/flash_attn_npu_4/fag_tiling.h
csrc/ascend910/flash_attn_npu_4/flash_api.cpp
csrc/ascend910/flash_attn_npu_4/fwd_dispatch.hpp
csrc/ascend910/flash_attn_npu_4/fwd_dispatch_impl.hpp
csrc/ascend910/flash_attn_npu_4/init_outputs.hpp
csrc/ascend910/flash_attn_npu_4/kernel_common.hpp
csrc/ascend910/flash_attn_npu_4/mha_fwd_kvcache.cpp
csrc/ascend910/flash_attn_npu_4/online_softmax.hpp
csrc/ascend910/flash_attn_npu_4/online_softmax_low_prec.hpp
csrc/ascend910/flash_attn_npu_4/pv_matmul.hpp
csrc/ascend910/flash_attn_npu_4/qk_matmul.hpp
csrc/ascend910/flash_attn_npu_4/rescale_o.hpp
csrc/ascend910/flash_attn_npu_4/rescale_o_low_prec.hpp
csrc/ascend910/flash_attn_npu_4/tilingdata.h
csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_bf16_bsnd.cpp
csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_bf16_tnd.cpp
csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_fp16_bsnd.cpp
csrc/ascend910/flash_attn_npu_4/autogen/bwd_dispatch_fp16_tnd.cpp
csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_bf16_bsnd.cpp
csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_bf16_tnd.cpp
csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_fp16_bsnd.cpp
csrc/ascend910/flash_attn_npu_4/autogen/fwd_dispatch_fp16_tnd.cpp
csrc/ascend910/flash_attn_npu_4/fag_common/epilogue_fag_deterministic_add.hpp
csrc/ascend910/flash_attn_npu_4/fag_common/mmad_fag_dqkv.hpp
csrc/ascend910/flash_attn_npu_4/fag_common/mmad_fag_sdp.hpp
csrc/ascend950/flash_attn_npu_3/fai_host_api.cpp
csrc/ascend950/flash_attn_npu_3/fai_host_api.hpp
csrc/ascend950/flash_attn_npu_3/fai_host_api_impl.hpp
csrc/ascend950/flash_attn_npu_3/fai_kernel.cpp
csrc/ascend950/flash_attn_npu_3/fai_tiling.cpp
csrc/ascend950/flash_attn_npu_3/fai_tilingdata.h
csrc/ascend950/flash_attn_npu_3/flash_api.cpp
csrc/ascend950/flash_attn_npu_3/kernel_common.hpp
csrc/ascend950/flash_attn_npu_3/mha_fwd.cpp
csrc/ascend950/flash_attn_npu_3/online_softmax.hpp
csrc/ascend950/flash_attn_npu_3/pv_matmul.hpp
csrc/ascend950/flash_attn_npu_3/qk_matmul.hpp
csrc/ascend950/flash_attn_npu_3/qk_matmul_DN.hpp
csrc/ascend950/flash_attn_npu_3/rescale_o.hpp
csrc/ascend950/flash_attn_npu_3/tiling_from_tensors.hpp
csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_bf16_bsnd.cpp
csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_bf16_tnd.cpp
csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_fp16_bsnd.cpp
csrc/ascend950/flash_attn_npu_3/autogen/fai_dispatch_fp16_tnd.cpp
csrc/ascend950/flash_attn_npu_4/fai_host_api.cpp
csrc/ascend950/flash_attn_npu_4/fai_host_api.hpp
csrc/ascend950/flash_attn_npu_4/fai_host_api_impl.hpp
csrc/ascend950/flash_attn_npu_4/fai_kernel.cpp
csrc/ascend950/flash_attn_npu_4/fai_tiling.cpp
csrc/ascend950/flash_attn_npu_4/fai_tilingdata.h
csrc/ascend950/flash_attn_npu_4/flash_api.cpp
csrc/ascend950/flash_attn_npu_4/kernel_common.hpp
csrc/ascend950/flash_attn_npu_4/mha_fwd.cpp
csrc/ascend950/flash_attn_npu_4/online_softmax_high_prec.hpp
csrc/ascend950/flash_attn_npu_4/online_softmax_low_prec.hpp
csrc/ascend950/flash_attn_npu_4/pv_matmul.hpp
csrc/ascend950/flash_attn_npu_4/qk_matmul.hpp
csrc/ascend950/flash_attn_npu_4/qk_matmul_DN.hpp
csrc/ascend950/flash_attn_npu_4/rescale_o.hpp
csrc/ascend950/flash_attn_npu_4/tiling_from_tensors.hpp
csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_bf16_bsnd.cpp
csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_bf16_tnd.cpp
csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_fp16_bsnd.cpp
csrc/ascend950/flash_attn_npu_4/autogen/fai_dispatch_fp16_tnd.cpp
csrc/catlass/include/catlass/catlass.hpp
csrc/catlass/include/catlass/conv_coord.hpp
csrc/catlass/include/catlass/coord.hpp
csrc/catlass/include/catlass/gemm_coord.hpp
csrc/catlass/include/catlass/gemv_coord.hpp
csrc/catlass/include/catlass/matrix_coord.hpp
csrc/catlass/include/catlass/numeric_size.hpp
csrc/catlass/include/catlass/status.hpp
csrc/catlass/include/catlass/arch/arch.hpp
csrc/catlass/include/catlass/arch/cross_core_sync.hpp
csrc/catlass/include/catlass/arch/local_tensor_buffer.hpp
csrc/catlass/include/catlass/arch/resource.hpp
csrc/catlass/include/catlass/conv/dispatch_policy.hpp
csrc/catlass/include/catlass/conv/block/block_conv.hpp
csrc/catlass/include/catlass/conv/block/block_conv2d_pingpong.hpp
csrc/catlass/include/catlass/conv/block/block_conv2d_pingpong_tla.hpp
csrc/catlass/include/catlass/conv/block/block_conv3d_pingpong_bias.hpp
csrc/catlass/include/catlass/conv/block/block_swizzle.hpp
csrc/catlass/include/catlass/conv/device/device_conv.hpp
csrc/catlass/include/catlass/conv/kernel/basic_conv2d.hpp
csrc/catlass/include/catlass/conv/kernel/basic_conv2d_tla.hpp
csrc/catlass/include/catlass/conv/kernel/conv3d_bias.hpp
csrc/catlass/include/catlass/conv/tile/copy_gm_to_l1.hpp
csrc/catlass/include/catlass/conv/tile/copy_l0c_to_gm.hpp
csrc/catlass/include/catlass/conv/tile/copy_l1_to_l0a.hpp
csrc/catlass/include/catlass/conv/tile/copy_l1_to_l0b.hpp
csrc/catlass/include/catlass/conv/tile/tile_copy.hpp
csrc/catlass/include/catlass/conv/tile/ascend950/copy_gm_to_l1.hpp
csrc/catlass/include/catlass/conv/tile/ascend950/copy_l0c_to_dst.hpp
csrc/catlass/include/catlass/conv/tile/ascend950/copy_l0c_to_gm.hpp
csrc/catlass/include/catlass/conv/tile/ascend950/copy_l1_to_l0a.hpp
csrc/catlass/include/catlass/conv/tile/ascend950/copy_l1_to_l0b.hpp
csrc/catlass/include/catlass/conv/tile/atlasa2/copy_gm_to_l1.hpp
csrc/catlass/include/catlass/conv/tile/atlasa2/copy_l0c_to_gm.hpp
csrc/catlass/include/catlass/conv/tile/atlasa2/copy_l1_to_l0a.hpp
csrc/catlass/include/catlass/conv/tile/atlasa2/copy_l1_to_l0b.hpp
csrc/catlass/include/catlass/detail/alignment.hpp
csrc/catlass/include/catlass/detail/callback.hpp
csrc/catlass/include/catlass/detail/dependent_false.hpp
csrc/catlass/include/catlass/detail/kernel_adapter.hpp
csrc/catlass/include/catlass/detail/macros.hpp
csrc/catlass/include/catlass/detail/tag_to_layout.hpp
csrc/catlass/include/catlass/epilogue/dispatch_policy.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_amla_tp1_rescale_o.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_amla_tp1_softmax.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_dequant.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_dual_level_quant_mx.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_elemwise_no_source.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_elemwise_one_source.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_fa_rescale_o.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_fa_rescale_o_ascend950.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_fa_softmax.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_fa_softmax_ascend950.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_finalize_routing.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_fixpipe.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_flash_attention_online_softmax_high_prec.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_flash_attention_online_softmax_low_prec.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_flash_attention_rescale_o.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_gemm.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_gemv.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_mla_fd_rescale_o.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_mla_rescale_o.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_mla_softmax.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_mla_tp1_rescale_o.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_mla_tp1_softmax.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_online_softmax_no_mask.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_per_block_quant_tla.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_per_group_per_block.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_per_token_dequant.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_per_token_dequant_tla.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_rescale_o_no_split_row.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_swiglu_mx_quant.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_visitor.hpp
csrc/catlass/include/catlass/epilogue/block/block_epilogue_w4a4_per_token_per_channel_dequant.hpp
csrc/catlass/include/catlass/epilogue/fusion/fusion.hpp
csrc/catlass/include/catlass/epilogue/fusion/operations.hpp
csrc/catlass/include/catlass/epilogue/fusion/topological_visitor.hpp
csrc/catlass/include/catlass/epilogue/fusion/tree_visitor.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_acc_load.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_aux_load.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_aux_store.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_cast.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_compute.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_impl.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_impl_base.hpp
csrc/catlass/include/catlass/epilogue/fusion/visitor_row_broadcast.hpp
csrc/catlass/include/catlass/epilogue/tile/copy_gm_to_ub.hpp
csrc/catlass/include/catlass/epilogue/tile/copy_gm_to_ub_tla.hpp
csrc/catlass/include/catlass/epilogue/tile/copy_ub_to_gm.hpp
csrc/catlass/include/catlass/epilogue/tile/copy_ub_to_gm_tla.hpp
csrc/catlass/include/catlass/epilogue/tile/copy_ub_to_l1_tla.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_broadcast_add.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_broadcast_inplace_by_column.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_broadcast_inplace_by_row.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_broadcast_mul.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_broadcast_one_blk.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_cast.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_copy.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_copy_dual_level_quant_mx.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_elemwise_add.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_elemwise_gelu.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_elemwise_mul.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_elemwise_muls.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_elemwise_silu.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_perblock_quant.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_pertoken_dequant.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_swiglu_mx_quant.hpp
csrc/catlass/include/catlass/epilogue/tile/tile_swizzle.hpp
csrc/catlass/include/catlass/gemm/dispatch_policy.hpp
csrc/catlass/include/catlass/gemm/gemm_type.hpp
csrc/catlass/include/catlass/gemm/helper.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_amla_pv_tp1_spec.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_dynamic_aiv.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_dynamic_common.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_dynamic_preload_async_with_callback.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_dynamic_single_core_splitk.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_dynamic_small.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_dynamic_streamk.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fa_pv.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fa_qk.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_pv_head_tail.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_pv_head_tail_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_pv_mx_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_pv_normal.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_pv_normal_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_pv_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_qk_head_tail.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_qk_head_tail_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_qk_mx_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_qk_normal.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_qk_normal_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_fai_qk_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_flash_attention_pv.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_flash_attention_qk.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_flash_attention_qk_DN.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_gemm.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_mla_pv.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_mla_pv_tp1_spec.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_mla_qk.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_mla_qk_tp1_spec.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_mx_a8w4.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_mx_preload_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_mx_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_bias.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_dequant_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_full_loadA.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_full_loadA_ascend950_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_full_loadA_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_mutex_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_per_group_per_block_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_slice_k_with_prologue.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_tla_v2.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_pingpong_with_prologue.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_preload.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_preload_async.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_preload_async_with_callback.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_preload_async_with_callback_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_preload_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_single_core_splitk.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_small.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_sparse_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_streamk.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_svd_quant_tla.hpp
csrc/catlass/include/catlass/gemm/block/block_mmad_w4a4_per_token_per_channel_dequant.hpp
csrc/catlass/include/catlass/gemm/block/block_mx_a8w4_prologue.hpp
csrc/catlass/include/catlass/gemm/block/block_scheduler_aswt.hpp
csrc/catlass/include/catlass/gemm/block/block_scheduler_iterateK.hpp
csrc/catlass/include/catlass/gemm/block/block_swizzle.hpp
csrc/catlass/include/catlass/gemm/block/block_swizzle_grouped_aswt.hpp
csrc/catlass/include/catlass/gemm/device/device_gemm.hpp
csrc/catlass/include/catlass/gemm/kernel/a8w4_mx_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/basic_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/basic_matmul_preload.hpp
csrc/catlass/include/catlass/gemm/kernel/basic_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/basic_matmul_tla_ub_visitor.hpp
csrc/catlass/include/catlass/gemm/kernel/basic_matmul_tla_visitor.hpp
csrc/catlass/include/catlass/gemm/kernel/batched_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/batched_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/broadcast_matmul_perblock_quant_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/dual_level_quant_mx_batched_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_aiv_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_common_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_local_padding_c_padding_common_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_padding_common_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_padding_multi_core_splitk_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_padding_streamk_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_single_core_splitk_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_small_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/dynamic_w8a8_matmul_per_token_per_channel_dequant.hpp
csrc/catlass/include/catlass/gemm/kernel/fp8_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/gemm.hpp
csrc/catlass/include/catlass/gemm/kernel/group_gemm.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_k.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_k_per_token_dequant.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_m.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_m_per_tensor_per_channel_dequant_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_m_per_token_dequant.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_m_per_token_dequant_multistage_workspace.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_m_per_token_dequant_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_matmul_slice_m_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_mx_matmul_finalize_routing_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_mx_matmul_slice_m_aswt_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_mx_matmul_slice_m_swiglu_mx_quant_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/grouped_mx_matmul_slice_m_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/matmul_activation.hpp
csrc/catlass/include/catlass/gemm/kernel/matmul_bias.hpp
csrc/catlass/include/catlass/gemm/kernel/matmul_epilogue.hpp
csrc/catlass/include/catlass/gemm/kernel/matmul_full_dequant.hpp
csrc/catlass/include/catlass/gemm/kernel/matmul_full_loadA.hpp
csrc/catlass/include/catlass/gemm/kernel/matmul_full_loadA_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/matmul_mix_fixpipe_opti.hpp
csrc/catlass/include/catlass/gemm/kernel/multi_core_splitk_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/mx_batched_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/mx_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/optimized_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/optimized_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/padding_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/padding_splitk_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/quant_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/quant_matmul_full_loadA_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/quant_matmul_multistage_workspace.hpp
csrc/catlass/include/catlass/gemm/kernel/quant_matmul_per_group_per_block_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/quant_multi_core_splitk_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/quant_optimized_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/single_core_slicek_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/small_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/sparse_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/splitk_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/streamk_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/streamk_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/strided_batched_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/svd_quant_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/tail_multi_core_splitk_matmul_tla.hpp
csrc/catlass/include/catlass/gemm/kernel/w4a4_matmul_per_token_per_channel_dequant.hpp
csrc/catlass/include/catlass/gemm/kernel/w4a8_matmul.hpp
csrc/catlass/include/catlass/gemm/kernel/w8a16_matmul.hpp
csrc/catlass/include/catlass/gemm/tile/cast_fp8_to_fp16.hpp
csrc/catlass/include/catlass/gemm/tile/cast_int4_to_int8.hpp
csrc/catlass/include/catlass/gemm/tile/cast_int8_to_fp16.hpp
csrc/catlass/include/catlass/gemm/tile/copy_gm_to_l1.hpp
csrc/catlass/include/catlass/gemm/tile/copy_gm_to_ub.hpp
csrc/catlass/include/catlass/gemm/tile/copy_l0c_to_gm.hpp
csrc/catlass/include/catlass/gemm/tile/copy_l0c_to_ub.hpp
csrc/catlass/include/catlass/gemm/tile/copy_l1_to_bt.hpp
csrc/catlass/include/catlass/gemm/tile/copy_l1_to_fp.hpp
csrc/catlass/include/catlass/gemm/tile/copy_l1_to_l0a.hpp
csrc/catlass/include/catlass/gemm/tile/copy_l1_to_l0b.hpp
csrc/catlass/include/catlass/gemm/tile/copy_ub_to_gm.hpp
csrc/catlass/include/catlass/gemm/tile/tile_copy.hpp
csrc/catlass/include/catlass/gemm/tile/tile_copy_tla.hpp
csrc/catlass/include/catlass/gemm/tile/tile_mmad.hpp
csrc/catlass/include/catlass/gemm/tile/tile_muls.hpp
csrc/catlass/include/catlass/gemm/tile/tile_traits.hpp
csrc/catlass/include/catlass/gemm/tile/ascend950/copy_gm_to_l1.hpp
csrc/catlass/include/catlass/gemm/tile/ascend950/copy_l0c_to_dst.hpp
csrc/catlass/include/catlass/gemm/tile/ascend950/copy_l0c_to_gm.hpp
csrc/catlass/include/catlass/gemm/tile/ascend950/copy_l0c_to_ub.hpp
csrc/catlass/include/catlass/gemm/tile/ascend950/copy_l1_to_bt.hpp
csrc/catlass/include/catlass/gemm/tile/ascend950/copy_l1_to_l0a.hpp
csrc/catlass/include/catlass/gemm/tile/ascend950/copy_l1_to_l0b.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/cast_fp8_to_fp16.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/cast_int4_to_int8.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/cast_int8_to_fp16.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/copy_gm_to_l1.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/copy_gm_to_ub.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/copy_l0c_to_gm.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/copy_l1_to_bt.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/copy_l1_to_l0a.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/copy_l1_to_l0b.hpp
csrc/catlass/include/catlass/gemm/tile/atlasa2/copy_ub_to_gm.hpp
csrc/catlass/include/catlass/gemv/helper.hpp
csrc/catlass/include/catlass/gemv/block/block_gemv.hpp
csrc/catlass/include/catlass/gemv/block/block_gemv_aic.hpp
csrc/catlass/include/catlass/gemv/block/block_gemv_aiv.hpp
csrc/catlass/include/catlass/gemv/device/device_gemv.hpp
csrc/catlass/include/catlass/gemv/kernel/kernel_gemv_aic.hpp
csrc/catlass/include/catlass/gemv/kernel/kernel_gemv_aiv.hpp
csrc/catlass/include/catlass/gemv/tile/matrix_copy_gm_to_ub.hpp
csrc/catlass/include/catlass/gemv/tile/tile_copy.hpp
csrc/catlass/include/catlass/gemv/tile/tile_vmad.hpp
csrc/catlass/include/catlass/gemv/tile/tile_vmuls.hpp
csrc/catlass/include/catlass/gemv/tile/vec_copy_gm_to_ub.hpp
csrc/catlass/include/catlass/gemv/tile/vec_copy_ub_to_gm.hpp
csrc/catlass/include/catlass/layout/layout.hpp
csrc/catlass/include/catlass/layout/matrix.hpp
csrc/catlass/include/catlass/layout/tensor.hpp
csrc/catlass/include/catlass/layout/vector.hpp
csrc/catlass/include/tla/int_tuple.hpp
csrc/catlass/include/tla/layout.hpp
csrc/catlass/include/tla/tensor.hpp
csrc/catlass/include/tla/tuple.hpp
csrc/catlass/include/tla/type_traits.hpp
csrc/catlass/include/tla/numeric/integer_sequence.hpp
csrc/catlass/include/tla/numeric/integral_constant.hpp
csrc/catlass/include/tla/numeric/math.hpp
csrc/catlass/tools/library/include/catlass/library/manifest.h
csrc/catlass/tools/library/include/catlass/library/operation.h
csrc/catlass/tools/library/src/gemm_operation.h
csrc/catlass/tools/library/src/library_utils.h
csrc/catlass/tools/library/src/manifest.cpp
flash_attn_npu/__init__.py
flash_attn_npu/flash_attn_npu_interface.py
flash_attn_npu.egg-info/PKG-INFO
flash_attn_npu.egg-info/SOURCES.txt
flash_attn_npu.egg-info/dependency_links.txt
flash_attn_npu.egg-info/requires.txt
flash_attn_npu.egg-info/top_level.txt
flash_attn_npu_3/__init__.py
flash_attn_npu_3/flash_attn_npu_interface.py
flash_attn_npu_3/flash_attn_npu_interface_950.py
flash_attn_npu_4/__init__.py
flash_attn_npu_4/flash_attn_npu_interface.py
flash_attn_npu_4/flash_attn_npu_interface_950.py
tests/test_flash_attn_npu_v2.py
tests/test_flash_attn_npu_v2_bwd.py
tests/test_flash_attn_npu_v3.py
tests/test_flash_attn_npu_v3_bwd.py
tests/test_flash_attn_npu_v3_graph.py
tests/test_flash_attn_npu_v3_metadata.py
tests/test_flash_attn_npu_v4.py
tests/test_flash_attn_npu_v4_bwd.py