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

↓ 1 callersMethodappend

(self, tile_size)

scripts/autogen_blackwell_fmha.py:349

↓ 1 callersMethodappend

(self, dim: int)

scripts/autogen_hopper_fmha_bwd.py:314

↓ 1 callersMethodappend

scripts/autogen_hopper_fmha_bwd.py:361

↓ 1 callersMethodappend

(self, dim: int)

scripts/autogen_hopper_fmha.py:348

↓ 1 callersMethodappend

scripts/autogen_hopper_fmha.py:393

↓ 1 callersMethodappend

(self, dtype: DataType)

scripts/autogen_hopper_fna.py:344

↓ 1 callersMethodappend

(self, dim: int)

scripts/autogen_hopper_fna.py:383

↓ 1 callersMethodappend

scripts/autogen_hopper_fna.py:428

↓ 1 callersMethodappend

scripts/autogen_hopper_fna.py:490

↓ 1 callersFunctionapply_variable_length

csrc/include/natten/cuda/fmha_blackwell/collective/fmha_fusion.hpp:315

↓ 1 callersFunctionapply_variable_length_offset

csrc/include/natten/cuda/fmha_blackwell/collective/fmha_fusion.hpp:345

↓ 1 callersFunctionarch_list_to_cmake_tags

(arch_list: List[int])

↓ 1 callersFunctionare_deterministic_algorithms_enabled

src/natten/context.py:109

↓ 1 callersFunctionatomicMaxFloat

csrc/include/natten/cuda/fmha/kernel_forward.h:64

↓ 1 callersFunctionatomicMaxFloat

csrc/include/natten/cuda/fna/kernel_forward.h:92

↓ 1 callersFunctionautogen_directories_match

Compare two directories recursively, ignoring timestamps and hidden files. Only compares files with .cu, .cpp, .h, .hpp, .cuh extensions.

↓ 1 callersFunctionautogen_kernel_instantitations

( this_dir: str, autogen_dir: str, scripts_dir: str, policy: str, cuda_arch_list: List[int

↓ 1 callersFunctionbold

src/natten/profiling_utils/pretty_printer.py:60

↓ 1 callersFunctioncheck_causal

( input_size: Sequence, causal: Optional[Sequence] = None )

src/natten/profiler.py:268

↓ 1 callersFunctioncheck_causal_arg

(na_dim: int, is_causal: Any)

src/natten/utils/checks.py:451

↓ 1 callersFunctioncheck_cutlass_blackwell_fmha_backward_config

( input_tensor: Tensor, q_tile_size: Optional[int] = None, kv_tile_size: Optional[int] = None, )

src/natten/backends/configs/cutlass_blackwell/__init__.py:348

↓ 1 callersFunctioncheck_cutlass_blackwell_fmha_forward_config

( input_tensor: Tensor, q_tile_size: Optional[int] = None, kv_tile_size: Optional[int] = None, )

src/natten/backends/configs/cutlass_blackwell/__init__.py:255

↓ 1 callersFunctioncheck_cutlass_blackwell_fna_backward_config

( input_tensor: Tensor, q_tile_shape: Optional[DimensionType] = None, kv_tile_shape: Optional[Dime

src/natten/backends/configs/cutlass_blackwell/__init__.py:301

↓ 1 callersFunctioncheck_cutlass_blackwell_fna_forward_config

( input_tensor: Tensor, q_tile_shape: Optional[DimensionType] = None, kv_tile_shape: Optional[Dime

src/natten/backends/configs/cutlass_blackwell/__init__.py:208

↓ 1 callersFunctioncheck_cutlass_fmha_backward_config

( input_tensor: Tensor, q_tile_size: Optional[int] = None, kv_tile_size: Optional[int] = None, )

src/natten/backends/configs/cutlass/__init__.py:385

↓ 1 callersFunctioncheck_cutlass_fmha_forward_config

( input_tensor: Tensor, q_tile_size: Optional[int] = None, kv_tile_size: Optional[int] = None, )

src/natten/backends/configs/cutlass/__init__.py:190

↓ 1 callersFunctioncheck_cutlass_fna_backward_config

( input_tensor: Tensor, q_tile_shape: Optional[DimensionType] = None, kv_tile_shape: Optional[Dime

src/natten/backends/configs/cutlass/__init__.py:335

↓ 1 callersFunctioncheck_cutlass_fna_forward_config

( input_tensor: Tensor, dilation: Optional[DimensionType] = None, q_tile_shape: Optional[Dimension

src/natten/backends/configs/cutlass/__init__.py:141

↓ 1 callersFunctioncheck_cutlass_hopper_fmha_backward_config

( input_tensor: Tensor, q_tile_size: Optional[int] = None, kv_tile_size: Optional[int] = None, )

src/natten/backends/configs/cutlass_hopper/__init__.py:475

↓ 1 callersFunctioncheck_cutlass_hopper_fmha_forward_config

( input_tensor: Tensor, q_tile_size: Optional[int] = None, kv_tile_size: Optional[int] = None,

src/natten/backends/configs/cutlass_hopper/__init__.py:420

↓ 1 callersFunctioncheck_cutlass_hopper_fna_backward_config

( input_tensor: Tensor, q_tile_shape: Optional[DimensionType] = None, kv_tile_shape: Optional[Dime

src/natten/backends/configs/cutlass_hopper/__init__.py:366

↓ 1 callersFunctioncheck_cutlass_hopper_fna_forward_config

( input_tensor: Tensor, q_tile_shape: Optional[DimensionType] = None, kv_tile_shape: Optional[Dime

src/natten/backends/configs/cutlass_hopper/__init__.py:307

↓ 1 callersFunctioncheck_dilation

( input_size: Sequence, window_size: Sequence, dilation: Optional[Sequence] = None )

src/natten/profiler.py:212

↓ 1 callersFunctioncheck_flex_fmha_forward_config

( input_tensor: Tensor, q_tile_size: Optional[int] = None, kv_tile_size: Optional[int] = None, )

src/natten/backends/configs/flex/__init__.py:166

↓ 1 callersFunctioncheck_flex_fna_forward_config

( input_tensor: Tensor, q_tile_shape: Optional[DimensionType] = None, kv_tile_shape: Optional[Dime

src/natten/backends/configs/flex/__init__.py:118

↓ 1 callersFunctioncheck_fmha_kv_splits

( kv_splits: Optional[int], input_tensor: Tensor, kv_tile_size: int, deterministic: bool,

src/natten/backends/configs/cutlass/backward_knobs.py:177

↓ 1 callersFunctioncheck_fna_kv_splits

( kv_splits: Optional[DimensionType], input_tensor: Tensor, kv_tile_shape: DimensionType, dete

src/natten/backends/configs/cutlass/backward_knobs.py:204

↓ 1 callersFunctioncheck_input_size

(input_size: Sequence)

src/natten/profiler.py:175

↓ 1 callersFunctioncheck_kernel_size_arg

(na_dim: int, kernel_size: Any)

src/natten/utils/checks.py:391

↓ 1 callersFunctioncheck_stride

( input_size: Sequence, window_size: Sequence, dilation: Sequence, stride: Optional[Sequence]

src/natten/profiler.py:238

↓ 1 callersFunctioncheck_stride_arg

(na_dim: int, stride: Any)

src/natten/utils/checks.py:409

↓ 1 callersFunctioncheck_window_size

( input_size: Sequence, window_size: Optional[Sequence] = None )

src/natten/profiler.py:187

↓ 1 callersMethodclear_mask

Clears the predicate set efficiently

csrc/include/natten/cuda/fna/iterators/predicated_tile_iterator_residual_last.h:548

↓ 1 callersFunctioncompute_delta_pt

(out: torch.Tensor, d_out: torch.Tensor, dtype)

tests/test_compute_delta.py:45

↓ 1 callersFunctioncompute_natten_fmha_reference

( batch, heads, dim, dim_v, seqlen_q, seqlen_kv, is_causal, dtype=torch.float3

tests/test_fmha.py:149

↓ 1 callersMethodcompute_predicates_

Computes predicates based on internally tracked per-thread offset.

csrc/include/natten/cuda/fmha/iterators/predicated_tile_access_iterator_residual_last.h:203

↓ 1 callersFunctioncompute_sdpa_reference

( batch, heads, heads_kv, dim, dim_v, seqlen_q, seqlen_kv, is_causal, dtyp

tests/test_fmha.py:78

↓ 1 callersFunctioncompute_split_reference

( batch: int, heads: int, head_dim: int, seqlens_Q_list: List[int], seqlens_KV_list: List[

tests/test_fmha_varlen.py:74

↓ 1 callersMethodcopy

csrc/include/natten/cuda/fmha_hopper/kernel/fmha_kernel_bwd_convert.hpp:120

↓ 1 callersFunctioncorrect_qkv_shape_wrt_dilation

csrc/include/natten/cuda/reference/mask.hpp:228

↓ 1 callersMethodcorrection

csrc/include/natten/cuda/fmha_blackwell/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp:1070

↓ 1 callersMethodcorrection

csrc/include/natten/cuda/fna_blackwell/collective/sm100_fna_fwd_mainloop_tma_warpspecialized.hpp:1150

↓ 1 callersMethodcorrection_empty

csrc/include/natten/cuda/fmha_blackwell/collective/sm100_fmha_fwd_mainloop_tma_warpspecialized.hpp:1297

↓ 1 callersMethodcorrection_empty

csrc/include/natten/cuda/fna_blackwell/collective/sm100_fna_fwd_mainloop_tma_warpspecialized.hpp:1378

↓ 1 callersFunctioncutlass_blackwell_fmha

( query: Tensor, key: Tensor, value: Tensor, is_causal: bool = False, scale: Optional[floa

src/natten/backends/blackwell_fmha.py:177

↓ 1 callersFunctioncutlass_fmha

( query: Tensor, key: Tensor, value: Tensor, is_causal: bool = False, scale: Optional[floa

src/natten/backends/fmha.py:188

↓ 1 callersFunctioncutlass_hopper_fmha

( query: Tensor, key: Tensor, value: Tensor, is_causal: bool = False, scale: Optional[floa

src/natten/backends/hopper_fmha.py:169

↓ 1 callersMethodenable

< CUTLASS_HOST_DEVICE enables all accesses guarded by mask

csrc/include/natten/cuda/fmha/iterators/epilogue_predicated_tile_iterator.h:170

↓ 1 callersMethodenable

< CUTLASS_HOST_DEVICE enables all accesses guarded by mask

csrc/include/natten/cuda/fna/iterators/epilogue_predicated_tile_iterator.h:179

↓ 1 callersMethodenable

< CUTLASS_HOST_DEVICE enables all accesses guarded by mask

csrc/include/natten/cuda/fna/epilogue/predicated_tile_iterator.h:176

↓ 1 callersMethodenable_mask

Clears the predicate set efficiently

csrc/include/natten/cuda/fna/iterators/predicated_tile_iterator_residual_last.h:559

↓ 1 callersFunctionfind_configs

( problem: Problem, backend: Optional[str], fmha_backend: Optional[str], backprop: bool, t

src/natten/profiling_utils/dry_run.py:328

↓ 1 callersFunctionflex_fmha

( query: Tensor, key: Tensor, value: Tensor, is_causal: bool = False, scale: Optional[floa

src/natten/backends/flex.py:179

↓ 1 callersFunctionfloor_div_tuple

csrc/include/natten/cuda/reference/mask.hpp:195

↓ 1 callersFunctionfna_bwd_reference_dK

csrc/include/natten/cuda/reference/fna_reference_backward.hpp:477

↓ 1 callersFunctionfna_bwd_reference_dQ

csrc/include/natten/cuda/reference/fna_reference_backward.hpp:405

↓ 1 callersFunctionfna_bwd_reference_dV

csrc/include/natten/cuda/reference/fna_reference_backward.hpp:549

↓ 1 callersFunctionformat_progress

Format progress as colored breakdown: passed+failed+xfailed+skipped+unknown+not_run / total.

scripts/testing/monitor.py:121

↓ 1 callersFunctiongemm_reset_zero_acc

csrc/include/natten/cuda/fna_hopper/collective/fna_common.hpp:86

↓ 1 callersFunctiongemm_reset_zero_acc

csrc/include/natten/cuda/fmha_hopper/collective/fmha_common.hpp:43

↓ 1 callersFunctiongenerate_blackwell_fmha

(output_directory: str, num_splits: int)

scripts/autogen_blackwell_fmha_bwd.py:604

↓ 1 callersFunctiongenerate_blackwell_fmha

(output_directory: str, num_splits: int)

scripts/autogen_blackwell_fmha.py:588

↓ 1 callersFunctiongenerate_blackwell_fmha_kernels

(path, num_splits=2)

scripts/autogen_blackwell_fmha_bwd.py:475

↓ 1 callersFunctiongenerate_blackwell_fmha_kernels

(path, num_splits=2)

scripts/autogen_blackwell_fmha.py:454

↓ 1 callersFunctiongenerate_blackwell_fna

(output_directory: str, num_splits: int)

scripts/autogen_blackwell_fna.py:774

↓ 1 callersFunctiongenerate_blackwell_fna

(output_directory: str, num_splits: int)

scripts/autogen_blackwell_fna_bwd.py:771

↓ 1 callersFunctiongenerate_blackwell_fna_kernels

(path, num_splits=2)

scripts/autogen_blackwell_fna.py:573

↓ 1 callersFunctiongenerate_blackwell_fna_kernels

(path, num_splits=2)

scripts/autogen_blackwell_fna_bwd.py:572

↓ 1 callersFunctiongenerate_cuda_fused

(output_directory: str, num_splits: int)

scripts/autogen_fna.py:885

↓ 1 callersFunctiongenerate_cuda_fused

(output_directory: str, num_splits: int)

scripts/autogen_fmha.py:655

↓ 1 callersFunctiongenerate_cuda_kernels

(path, num_splits=2)

scripts/autogen_fna.py:556

↓ 1 callersFunctiongenerate_cuda_kernels

(path, num_splits=2)

scripts/autogen_fmha.py:393

↓ 1 callersFunctiongenerate_hopper_fmha

(output_directory: str, num_splits: int)

scripts/autogen_hopper_fmha_bwd.py:601

↓ 1 callersFunctiongenerate_hopper_fmha

(output_directory: str, num_splits: int)

scripts/autogen_hopper_fmha.py:650

↓ 1 callersFunctiongenerate_hopper_fmha_kernels

(path, num_splits=2)

scripts/autogen_hopper_fmha_bwd.py:460

↓ 1 callersFunctiongenerate_hopper_fmha_kernels

(path, num_splits=2)

scripts/autogen_hopper_fmha.py:504

↓ 1 callersFunctiongenerate_hopper_fna

(output_directory: str, num_splits: int)

scripts/autogen_hopper_fna_bwd.py:819

↓ 1 callersFunctiongenerate_hopper_fna

(output_directory: str, num_splits: int)

scripts/autogen_hopper_fna.py:866

↓ 1 callersFunctiongenerate_hopper_fna_kernels

(path, num_splits=2)

scripts/autogen_hopper_fna_bwd.py:581

↓ 1 callersFunctiongenerate_hopper_fna_kernels

(path, num_splits=2)

scripts/autogen_hopper_fna.py:623

↓ 1 callersFunctiongenerate_problem

( batch_size: int, heads: int, input_size: DimensionType, dim: int, window_size: Dimension

src/natten/profiling_utils/problem.py:228

↓ 1 callersFunctiongenerate_reference_fna

(output_directory: str, num_splits: int)

scripts/autogen_reference_fna.py:602

↓ 1 callersFunctiongenerate_reference_fna_kernels

(path, num_splits=2)

scripts/autogen_reference_fna.py:468

↓ 1 callersMethodget_add_k_tensor_shape

(self, heads_last: bool)

src/natten/profiling_utils/problem.py:118

↓ 1 callersMethodget_add_v_tensor_shape

(self, heads_last: bool)

src/natten/profiling_utils/problem.py:124

↓ 1 callersFunctionget_all_fmha_backward_configs

( input_tensor: Tensor, )

src/natten/backends/configs/cutlass_blackwell/__init__.py:146

↓ 1 callersFunctionget_all_fmha_forward_configs

( input_tensor: Tensor, )

src/natten/backends/configs/cutlass_blackwell/__init__.py:132

↓ 1 callersFunctionget_all_tile_sizes_backward

( input_tensor: Tensor, )

src/natten/backends/configs/cutlass/__init__.py:305

↓ 1 callersFunctionget_all_tile_sizes_forward

(input_tensor: Tensor)

src/natten/backends/configs/flex/__init__.py:89

↓ 1 callersFunctionget_all_tile_sizes_forward

( input_tensor: Tensor, )

src/natten/backends/configs/cutlass/__init__.py:115

← previous next →501–600 of 2,059, ranked by callers