Functions — github.com/SHI-Labs/NATTEN

CodeHub MCPcopy Create free account

hub / github.com/SHI-Labs/NATTEN / functions

Functions2,059 in github.com/SHI-Labs/NATTEN

⨍ Functions 2,059◇ Types & classes 573

Methodadd_pointer_offset

Adds a pointer offset in units of Element

csrc/include/natten/cuda/fna/iterators/predicated_tile_iterator.h:503

Methodadd_pointer_offset

Adds a pointer offset in units of Element

csrc/include/natten/cuda/fna/iterators/epilogue_predicated_tile_iterator.h:290

Methodadd_pointer_offset

Adds a pointer offset in units of Element

csrc/include/natten/cuda/fna/iterators/predicated_tile_iterator_residual_last.h:232

Functionadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fna/iterators/predicated_tile_access_iterator_residual_last.h:303

Methodadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fmha/iterators/predicated_tile_access_iterator_residual_last.h:600

Methodadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fmha/iterators/predicated_tile_access_iterator_residual_last.h:1375

Methodadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fmha/iterators/predicated_tile_access_iterator_residual_last.h:1594

Methodadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fmha/iterators/predicated_tile_access_iterator_residual_last.h:1822

Methodadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fmha/iterators/predicated_tile_access_iterator_residual_last.h:2049

Methodadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fna/iterators/predicated_tile_access_iterator_residual_last.h:905

Methodadd_tile_offset

Advances an iterator along logical dimensions of matrix in units of whole tiles

csrc/include/natten/cuda/fna/iterators/predicated_tile_access_iterator.h:1368

Methodadvance_to_block

Moves pointers to what we should process Returns "false" if there is no work to do

csrc/include/natten/cuda/fmha/kernel_forward.h:173

Methodadvance_to_block

csrc/include/natten/cuda/fmha/kernel_backward.h:689

Methodadvance_to_block

Moves pointers to what we should process Returns "false" if there is no work to do

csrc/include/natten/cuda/fna/kernel_forward.h:205

Methodadvance_to_block

csrc/include/natten/cuda/fna/kernel_backward.h:737

Functionall_dims_match

csrc/include/natten/natten.h:80

Functionallow_flex_compile_backprop

Sets guards for Flex Attention + `torch.compile` for backpropagation only. Args: mode: If `True`, enable compilation for backprop (assumi

src/natten/context.py:214

Functionany_true

csrc/include/natten/natten.h:43

csrc/include/natten/cuda/fmha/gemm_kernel_utils.h:194

csrc/include/natten/cuda/fmha/gemm_kernel_utils.h:203

cast scale_frag to correct type then apply elementwise to fragment

csrc/include/natten/cuda/fmha/gemm/mma_from_smem.h:314

csrc/include/natten/cuda/fmha/gemm/mma_from_smem.h:329

csrc/include/natten/cuda/fmha/epilogue/epilogue_rescale_output.h:246

csrc/include/natten/cuda/fmha/epilogue/epilogue_pipelined.h:85

csrc/include/natten/cuda/fna/gemm_kernel_utils.h:139

csrc/include/natten/cuda/fna/gemm_kernel_utils.h:148

cast scale_frag to correct type then apply elementwise to fragment

csrc/include/natten/cuda/fna/gemm/mma_from_smem.h:315

csrc/include/natten/cuda/fna/gemm/mma_from_smem.h:330

csrc/include/natten/cuda/fna/epilogue/epilogue_rescale_output.h:246

Methodapply_batch

csrc/include/natten/cuda/fmha_blackwell/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp:277

Methodapply_batch

csrc/include/natten/cuda/fna_blackwell/kernel/sm100_fna_fwd_kernel_tma_warpspecialized.hpp:234

Methodapply_extra_kv_mask

csrc/include/natten/cuda/fna_blackwell/collective/fna_fusion.hpp:478

Methodapply_mask

csrc/include/natten/cuda/fmha_blackwell/collective/fmha_fusion.hpp:102

Methodapply_mask

csrc/include/natten/cuda/fmha_blackwell/collective/fmha_fusion.hpp:148

Methodapply_mask

csrc/include/natten/cuda/fmha_blackwell/collective/fmha_fusion.hpp:227

Methodapply_mask

csrc/include/natten/cuda/fmha_blackwell/collective/fmha_fusion.hpp:267

Methodapply_mask

csrc/include/natten/cuda/fna_hopper/collective/fna_fusion_bwd.hpp:326

Methodapply_mask

csrc/include/natten/cuda/fna_blackwell/collective/fna_fusion_bwd.hpp:324

Methodapply_output_operator_

Helper to invoke the output functor over each vector of output

csrc/include/natten/cuda/fmha/epilogue/epilogue_pipelined.h:542

Methodapply_output_operator_

Helper to invoke the output functor over each vector of output

csrc/include/natten/cuda/fna/epilogue/epilogue_pipelined.h:542

Methodapply_output_operator_source_not_needed_

Helper to invoke the output functor over each vector of output

csrc/include/natten/cuda/fmha/epilogue/epilogue_pipelined.h:573

Methodapply_output_operator_source_not_needed_

Helper to invoke the output functor over each vector of output

csrc/include/natten/cuda/fna/epilogue/epilogue_pipelined.h:573

Methodapply_padded_mask

csrc/include/natten/cuda/fna_hopper/collective/fna_fusion_bwd.hpp:404

Methodapply_padded_mask

csrc/include/natten/cuda/fna_blackwell/collective/fna_fusion_bwd.hpp:401

Methodattention_kernel

csrc/include/natten/cuda/fmha/kernel_forward.h:542

Methodattention_kernel

csrc/include/natten/cuda/fmha/kernel_backward.h:1163

Methodattention_kernel

csrc/include/natten/cuda/fna/kernel_forward.h:574

Methodattention_kernel

csrc/include/natten/cuda/fna/kernel_backward.h:1245

(ctx, grad_out: Tensor, grad_lse: Tensor)

src/natten/attn_merge.py:172

(ctx, d_output: Tensor)

src/natten/token_permute/cutlass_impl.py:116

(ctx, d_output: Tensor)

src/natten/token_permute/cutlass_impl.py:172

(ctx, grad_out: Tensor, grad_lse: Tensor)

src/natten/backends/hopper_fmha.py:115

(ctx, grad_out: Tensor, grad_lse: Tensor)

src/natten/backends/reference.py:134

(ctx, grad_out: Tensor, grad_lse: Tensor)

src/natten/backends/blackwell_fmha.py:115

(ctx, d_output: Tensor, d_lse: Tensor)

src/natten/backends/blackwell_fna.py:185

(ctx, grad_out: Tensor, grad_lse: Tensor)

src/natten/backends/fna.py:155

(ctx, d_output: Tensor, d_lse: Tensor)

src/natten/backends/hopper_fna.py:185

Methodbefore_softmax

csrc/include/natten/cuda/fmha_hopper/collective/fmha_fusion.hpp:68

Methodbefore_softmax

csrc/include/natten/cuda/fmha_hopper/collective/fmha_fusion.hpp:98

Methodbefore_softmax

csrc/include/natten/cuda/fmha_hopper/collective/fmha_fusion.hpp:188

Methodbefore_softmax

csrc/include/natten/cuda/fmha_hopper/collective/fmha_fusion.hpp:246

Functionblackwell_fmha_backward_torch_fake_op

( query: Tensor, key: Tensor, value: Tensor, output: Tensor, d_output: Tensor, logsume

src/natten/_libnatten/torch_wrappers.py:254

Functionblackwell_fmha_backward_torch_op

( query: Tensor, key: Tensor, value: Tensor, output: Tensor, d_output: Tensor, logsume

src/natten/_libnatten/torch_wrappers.py:195

Functionblackwell_fmha_forward_torch_fake_op

( query: Tensor, key: Tensor, value: Tensor, is_causal: bool, scale: float, q_tile_siz

src/natten/_libnatten/torch_wrappers.py:166

Functionblackwell_fmha_forward_torch_op

( query: Tensor, key: Tensor, value: Tensor, is_causal: bool, scale: float, q_tile_siz

src/natten/_libnatten/torch_wrappers.py:112

Functionblackwell_fna_backward_torch_fake_op

( query: Tensor, key: Tensor, value: Tensor, output: Tensor, d_output:

src/natten/_libnatten/torch_wrappers.py:802

Functionblackwell_fna_backward_torch_op

( query: Tensor, key: Tensor, value: Tensor, output: Tensor, d_output:

src/natten/_libnatten/torch_wrappers.py:750

Functionblackwell_fna_forward_torch_fake_op

( query: Tensor, key: Tensor, value: Tensor, kernel_size: list[int], s

src/natten/_libnatten/torch_wrappers.py:717

Functionblackwell_fna_forward_torch_op

( query: Tensor, key: Tensor, value: Tensor, kernel_size: list[int], s

src/natten/_libnatten/torch_wrappers.py:670

Methodbuild_extension

Methodcan_implement

csrc/include/natten/cuda/fmha_blackwell/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp:248

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/fmha_blackwell/device/fmha_sm100.hpp:87

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/fmha_blackwell/device/fmha_bwd_sm100.hpp:276

Methodcan_implement

csrc/include/natten/cuda/fmha_blackwell/kernel/sm100_fmha_fwd_kernel_tma_warpspecialized.hpp:251

Methodcan_implement

csrc/include/natten/cuda/fmha_blackwell/kernel/fmha_kernel_bwd_convert.hpp:96

Methodcan_implement

csrc/include/natten/cuda/fmha_blackwell/kernel/sm100_fmha_bwd_kernel_tma_warpspecialized.hpp:466

Methodcan_implement

csrc/include/natten/cuda/fmha_blackwell/kernel/fmha_kernel_bwd_sum_OdO.hpp:99

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/utils/generic_cutlass_device.hpp:88

Methodcan_implement

csrc/include/natten/cuda/fna_hopper/collective/fna_collective_tma_warpspecialized.hpp:253

Methodcan_implement

csrc/include/natten/cuda/fna_hopper/collective/fna_collective_bwd_tma_warpspecialized.hpp:411

Methodcan_implement

csrc/include/natten/cuda/fna_hopper/collective/fna_collective_tma.hpp:233

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/fna_hopper/device/fna_bwd_sm90.hpp:259

Methodcan_implement

csrc/include/natten/cuda/fna_blackwell/collective/sm100_fna_fwd_mainloop_tma_warpspecialized.hpp:268

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/fna_blackwell/device/fna_sm100.hpp:87

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/fna_blackwell/device/fna_bwd_sm100.hpp:264

Methodcan_implement

csrc/include/natten/cuda/fna_blackwell/kernel/sm100_fna_bwd_kernel_tma_warpspecialized.hpp:464

Methodcan_implement

csrc/include/natten/cuda/fna_blackwell/kernel/sm100_fna_fwd_kernel_tma_warpspecialized.hpp:208

Methodcan_implement

csrc/include/natten/cuda/fmha_hopper/collective/fmha_collective_tma.hpp:199

Methodcan_implement

csrc/include/natten/cuda/fmha_hopper/collective/fmha_collective_tma_warpspecialized.hpp:219

Methodcan_implement

csrc/include/natten/cuda/fmha_hopper/collective/fmha_collective_bwd_tma_warpspecialized.hpp:367

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/fmha_hopper/device/fmha_sm90.hpp:87

Methodcan_implement

Determines whether the GEMM can execute the given problem.

csrc/include/natten/cuda/fmha_hopper/device/fmha_bwd_sm90.hpp:233

Methodcan_implement

csrc/include/natten/cuda/fmha_hopper/kernel/fmha_kernel_tma_warpspecialized.hpp:157

Methodcan_implement

csrc/include/natten/cuda/fmha_hopper/kernel/fmha_kernel_bwd_convert.hpp:92

Methodcan_implement

csrc/include/natten/cuda/fmha_hopper/kernel/fmha_kernel_tma.hpp:123

Methodcan_implement

csrc/include/natten/cuda/fmha_hopper/kernel/fmha_kernel_bwd_sum_OdO.hpp:101

Methodcan_implement

csrc/include/natten/cuda/reduction/fmha_kernel_bwd_sum_OdO.hpp:85

Methodcausal_mask_inst

scripts/autogen_fna.py:141

Functioncheck_additional_keys

( input_tensor: Tensor, additional_keys: Optional[Tensor] )

src/natten/utils/tensor.py:44

Functioncheck_additional_values

( attn_tensor: Tensor, additional_values: Optional[Tensor], value: Tensor, expected_attn_weigh

src/natten/utils/tensor.py:72

← previous next →1,001–1,100 of 2,059, ranked by callers