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 callersFunctionget_arch
(sym: str)
src/natten/profiling_utils/formatting.py:296
↓ 1 callersFunctionget_args
()
src/natten/profiler.py:280
↓ 1 callersFunctionget_bwd_configs_for_cutlass_blackwell_fmha
Returns Blackwell FMHA backward pass configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Blackwell data
src/natten/backends/configs/__init__.py:122
↓ 1 callersFunctionget_bwd_configs_for_cutlass_blackwell_fna
Returns Blackwell FNA backward pass configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Blackwell datac
src/natten/backends/configs/__init__.py:191
↓ 1 callersFunctionget_bwd_configs_for_cutlass_fmha
Returns CUTLASS FMHA backward pass configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a device with comp
src/natten/backends/configs/__init__.py:412
↓ 1 callersFunctionget_bwd_configs_for_cutlass_fna
Returns CUTLASS FNA backward pass configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a device with compu
src/natten/backends/configs/__init__.py:480
↓ 1 callersFunctionget_bwd_configs_for_cutlass_hopper_fmha
Returns Hopper FMHA backward pass configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Hopper GPU (SM90;
src/natten/backends/configs/__init__.py:263
↓ 1 callersFunctionget_bwd_configs_for_cutlass_hopper_fna
Returns Hopper FNA backward pass configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Hopper GPU (SM90;
src/natten/backends/configs/__init__.py:337
↓ 1 callersFunctionget_bwd_stride_offset
csrc/include/natten/cuda/fna_hopper/collective/fna_fusion_bwd.hpp:445
↓ 1 callersFunctionget_bwd_stride_offset
csrc/include/natten/cuda/fna_blackwell/collective/fna_fusion_bwd.hpp:445
↓ 1 callersMethodget_causal_cute
(self)
scripts/autogen_hopper_fna_bwd.py:183
↓ 1 callersMethodget_causal_cute
(self)
scripts/autogen_blackwell_fna.py:176
↓ 1 callersMethodget_causal_cute
(self)
scripts/autogen_blackwell_fna_bwd.py:183
↓ 1 callersMethodget_causal_cute
(self)
scripts/autogen_reference_fna.py:209
↓ 1 callersMethodget_causal_cute
(self)
scripts/autogen_hopper_fna.py:216
↓ 1 callersFunctionget_configs_for_cutlass_blackwell_fmha
Returns Blackwell FMHA configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Blackwell datacenter GPU (SM
src/natten/backends/configs/__init__.py:85
↓ 1 callersFunctionget_configs_for_cutlass_blackwell_fna
Returns Blackwell FNA configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Blackwell datacenter GPU (SM1
src/natten/backends/configs/__init__.py:159
↓ 1 callersFunctionget_configs_for_cutlass_fmha
Returns CUTLASS FMHA configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a device with compute capability
src/natten/backends/configs/__init__.py:376
↓ 1 callersFunctionget_configs_for_cutlass_fna
Returns CUTLASS FNA configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a device with compute capability
src/natten/backends/configs/__init__.py:450
↓ 1 callersFunctionget_configs_for_cutlass_hopper_fmha
Returns Hopper FMHA configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Hopper GPU (SM90; compute capab
src/natten/backends/configs/__init__.py:226
↓ 1 callersFunctionget_configs_for_cutlass_hopper_fna
Returns Hopper FNA configurations compatible with input tensors, if any. Checks first if a CUDA tensor, and on a Hopper GPU (SM90; compute capabi
src/natten/backends/configs/__init__.py:303
↓ 1 callersFunctionget_configs_for_flex_fmha
Returns Flex FMHA configurations compatible with input tensors, if any. Each configuration for this operation is a tuple of two integers: `(q_til
src/natten/backends/configs/__init__.py:515
↓ 1 callersFunctionget_configs_for_flex_fna
Returns Flex FNA configurations compatible with input tensors, if any. Each configuration for this operation is a tuple of two integer tuples: `(
src/natten/backends/configs/__init__.py:550
↓ 1 callersFunctionget_cuda_arch_list
(cuda_arch_list: str)
setup.py:158
↓ 1 callersMethodget_decl
(self)
scripts/autogen_hopper_fna_bwd.py:201
↓ 1 callersMethodget_decl
(self)
scripts/autogen_blackwell_fmha_bwd.py:227
↓ 1 callersMethodget_decl
(self)
scripts/autogen_blackwell_fna.py:193
↓ 1 callersMethodget_decl
(self)
scripts/autogen_blackwell_fna_bwd.py:198
↓ 1 callersMethodget_decl
(self)
scripts/autogen_reference_fna.py:222
↓ 1 callersMethodget_decl
(self)
scripts/autogen_blackwell_fmha.py:198
↓ 1 callersMethodget_decl
(self)
scripts/autogen_hopper_fmha_bwd.py:209
↓ 1 callersMethodget_decl
(self)
scripts/autogen_hopper_fmha.py:244
↓ 1 callersMethodget_decl
(self)
scripts/autogen_hopper_fna.py:235
↓ 1 callersFunctionget_default_backward_tile_sizes
( input_tensor: Tensor, )
src/natten/backends/configs/cutlass_blackwell/__init__.py:196
↓ 1 callersFunctionget_default_fmha_backward_config
( input_tensor: Tensor, )
src/natten/backends/configs/cutlass_hopper/__init__.py:295
↓ 1 callersFunctionget_default_fmha_forward_config
( input_tensor: Tensor, )
src/natten/backends/configs/cutlass_hopper/__init__.py:283
↓ 1 callersFunctionget_default_forward_tile_sizes
(input_tensor: Tensor)
src/natten/backends/configs/flex/__init__.py:108
↓ 1 callersFunctionget_default_forward_tile_sizes
( input_tensor: Tensor, )
src/natten/backends/configs/cutlass_blackwell/__init__.py:169
↓ 1 callersFunctionget_flex_attention_fn
( torch_compile: bool, torch_compile_args: Optional[dict] = None )
src/natten/backends/flex.py:76
↓ 1 callersFunctionget_framework
(sym: str)
src/natten/profiling_utils/formatting.py:304
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_hopper_fna_bwd.py:189
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_blackwell_fmha_bwd.py:218
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_blackwell_fna.py:173
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_blackwell_fna_bwd.py:180
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_blackwell_fmha.py:187
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_hopper_fmha_bwd.py:200
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_hopper_fmha.py:234
↓ 1 callersMethodget_gemm_shape_cute
(self)
scripts/autogen_hopper_fna.py:222
↓ 1 callersMethodget_impl
(self)
scripts/autogen_hopper_fna_bwd.py:206
↓ 1 callersMethodget_impl
(self)
scripts/autogen_blackwell_fmha_bwd.py:232
↓ 1 callersMethodget_impl
(self)
scripts/autogen_blackwell_fna.py:198
↓ 1 callersMethodget_impl
(self)
scripts/autogen_blackwell_fna_bwd.py:203
↓ 1 callersMethodget_impl
(self)
scripts/autogen_reference_fna.py:231
↓ 1 callersMethodget_impl
(self)
scripts/autogen_blackwell_fmha.py:203
↓ 1 callersMethodget_impl
(self)
scripts/autogen_hopper_fmha_bwd.py:214
↓ 1 callersMethodget_impl
(self)
scripts/autogen_hopper_fmha.py:249
↓ 1 callersMethodget_impl
(self)
scripts/autogen_hopper_fna.py:240
↓ 1 callersMethodget_k_tensor_shape
(self, heads_last: bool, flatten: bool = False)
src/natten/profiling_utils/problem.py:100
↓ 1 callersMethodget_kv_tile_shape_cute
(self)
scripts/autogen_hopper_fna_bwd.py:180
↓ 1 callersMethodget_kv_tile_shape_cute
(self)
scripts/autogen_blackwell_fna.py:170
↓ 1 callersMethodget_kv_tile_shape_cute
(self)
scripts/autogen_blackwell_fna_bwd.py:177
↓ 1 callersMethodget_kv_tile_shape_cute
(self)
scripts/autogen_hopper_fna.py:213
↓ 1 callersMethodget_mask
Gets the mask
csrc/include/natten/cuda/fna/iterators/predicated_tile_iterator_residual_last.h:571
↓ 1 callersFunctionget_na_flex_mask
( device: str, na_dim: int, qkv_shape: DimensionType, kernel_size: DimensionType, stride:
src/natten/backends/flex.py:282
↓ 1 callersFunctionget_natten_op
(sym: str)
src/natten/profiling_utils/formatting.py:205
↓ 1 callersMethodget_o_tensor_shape
(self, heads_last: bool, flatten: bool = False)
src/natten/profiling_utils/problem.py:113
↓ 1 callersFunctionget_op_prio
(op: Optional[LibNattenOp])
src/natten/profiling_utils/formatting.py:312
↓ 1 callersFunctionget_pointer_offset
csrc/include/natten/cuda/fna/iterators/predicated_tile_access_iterator_residual_last.h:325
↓ 1 callersFunctionget_pointer_offset
csrc/include/natten/cuda/fna/iterators/predicated_tile_access_iterator.h:770
↓ 1 callersMethodget_q_tensor_shape
(self, heads_last: bool, flatten: bool = False)
src/natten/profiling_utils/problem.py:95
↓ 1 callersMethodget_q_tile_shape_cute
(self)
scripts/autogen_hopper_fna_bwd.py:177
↓ 1 callersMethodget_q_tile_shape_cute
(self)
scripts/autogen_blackwell_fna.py:167
↓ 1 callersMethodget_q_tile_shape_cute
(self)
scripts/autogen_blackwell_fna_bwd.py:174
↓ 1 callersMethodget_q_tile_shape_cute
(self)
scripts/autogen_hopper_fna.py:210
↓ 1 callersMethodget_target_name
(self, q_tile_shape, kv_tile_shape)
scripts/autogen_hopper_fna_bwd.py:485
↓ 1 callersMethodget_target_name
(self, q_tile_size, kv_tile_size)
scripts/autogen_blackwell_fmha_bwd.py:399
↓ 1 callersMethodget_target_name
(self, q_tile_shape, kv_tile_shape)
scripts/autogen_blackwell_fna_bwd.py:477
↓ 1 callersMethodget_target_name
(self, cm)
scripts/autogen_reference_fna.py:394
↓ 1 callersMethodget_target_name
(self, q_tile_size, kv_tile_size)
scripts/autogen_hopper_fmha_bwd.py:385
↓ 1 callersMethodget_target_name
(self, q_tile_size, kv_tile_size, kernel_type)
scripts/autogen_hopper_fmha.py:420
↓ 1 callersMethodget_target_name
(self, q_tile_shape, kv_tile_shape, kernel_type)
scripts/autogen_hopper_fna.py:521
↓ 1 callersMethodget_unmasked_trip_count
csrc/include/natten/cuda/fmha_blackwell/collective/fmha_fusion.hpp:59
↓ 1 callersMethodget_v_tensor_shape
(self, heads_last: bool, flatten: bool = False)
src/natten/profiling_utils/problem.py:105
↓ 1 callersFunctionget_version
()
setup.py:120
↓ 1 callersFunctionget_window_left
csrc/include/natten/cuda/reference/mask.hpp:49
↓ 1 callersFunctionget_window_right
csrc/include/natten/cuda/reference/mask.hpp:54
↓ 1 callersFunctionis_cpu
(device: torch.device)
src/natten/utils/device.py:37
↓ 1 callersFunctionis_equal
csrc/include/natten/cuda/reference/mask.hpp:207
↓ 1 callersFunctionis_flex_compile_backprop_allowed
Returns whether compilation for backpropagation is allowed in `"flex-fna"` and `"flex-fmha"` backends.
src/natten/context.py:154
↓ 1 callersFunctionis_foreground
()
scripts/testing/monitor.py:278
↓ 1 callersFunctionis_kv_parallelism_in_fused_na_enabled
Returns whether KV parallelism in `"cutlass-fna"` and `"cutlass-fmha"` backends is enabled.
src/natten/context.py:144
↓ 1 callersFunctionis_memory_usage_strict
Returns whether memory usage preference for KV parallelism in `"cutlass-fna"` and `"cutlass-fmha"` backends is the *restricted* setting.
src/natten/context.py:85
↓ 1 callersFunctionis_memory_usage_unrestricted
Returns whether memory usage preference for KV parallelism in `"cutlass-fna"` and `"cutlass-fmha"` backends is the *unrestricted* setting.
src/natten/context.py:92
↓ 1 callersFunctionis_neighbor
csrc/include/natten/cuda/reference/mask.hpp:173
↓ 1 callersFunctionis_rocm
(device: torch.device)
src/natten/utils/device.py:33
↓ 1 callersFunctionis_self_attention
( input_tensor: Tensor, kernel_size: DimensionType, is_causal: CausalArgType, has_additional_a
src/natten/utils/checks.py:511
↓ 1 callersMethodis_source_needed
Returns true if source is needed
csrc/include/natten/cuda/fmha/epilogue/epilogue_thread_apply_logsumexp.h:140
↓ 1 callersMethodis_source_needed
Returns true if source is needed
csrc/include/natten/cuda/fna/epilogue/epilogue_thread_apply_logsumexp.h:140
↓ 1 callersFunctioniterable_to_static_cute_tuple
(shape_in)
scripts/autogen_blackwell_fmha_bwd.py:202
↓ 1 callersFunctioniterable_to_static_cute_tuple
(shape_in)
scripts/autogen_blackwell_fmha.py:169
← previousnext →601–700 of 2,059, ranked by callers