MCPcopy Create free account

hub / github.com/pytorch/executorch / functions

Functions28,493 in github.com/pytorch/executorch

↓ 16 callersMethodcreate_recipe
Create CoreML recipe with precision and compute unit combinations
backends/apple/coreml/recipes/coreml_recipe_provider.py:42
↓ 16 callersMethodexport
(self)
examples/qualcomm/oss_scripts/llm_utils/qnn_decoder_model_manager.py:158
↓ 16 callersMethodexport
Exports the model pre-autograd. This is not a full export, since it uses torch.export.export() to keep autograd-safe ops from getting
extension/llm/export/builder.py:230
↓ 16 callersFunctionformat_target_name
We remove the dialect name space from the target name. We generally do not care for the op dialect specific name space ("aten.", "quantized_d
exir/backend/canonical_partitioners/config_partitioner.py:31
↓ 16 callersMethodgenerate_compile_specs
Returns the list of compile specs that's used by CoreMLBackend to lower the module.
backends/apple/coreml/compiler/coreml_preprocess.py:362
↓ 16 callersMethodget_inputs
(self)
backends/arm/test/ops/test_transpose_conv2d.py:51
↓ 16 callersMethodget_placeholder_target_and_tensor
(self, node: Node)
backends/mlx/builder/program_builder.py:265
↓ 16 callersMethodrun_tester
(self, module, inputs)
backends/xnnpack/test/passes/test_channels_last_tagged_reshape.py:55
↓ 16 callersFunctionserialize_from_bundled_program_to_flatbuffer
Serialize a BundledProgram into FlatBuffer binary format. Args: bundled_program (BundledProgram): The `BundledProgram` variable to b
devtools/bundled_program/serialize/__init__.py:82
↓ 16 callersFunctionserialize_pte_binary
Returns the runtime binary representation of the given Program. Args: pte_file: PTEFile class containing the program and segments.
exir/_serialize/_program.py:422
↓ 16 callersMethodset_module_type
Set quantization_config for submodules with a given module type. For example, calling set_module_type(Softmax) quantizes supported pa
backends/arm/quantizer/arm_quantizer.py:596
↓ 16 callersMethodtransform
Returns a transformed ExportedProgram
backends/xnnpack/_passes/__init__.py:96
↓ 16 callersFunctiontransposition_is_supported_on_neutron
This function determines if the current NeutronSoftware properly supports a `Transpose` operator with given `input_shape` and `permutation`.
backends/nxp/backend/neutron_operator_support.py:26
↓ 15 callersMethodBuiltinOptionsType
(self)
backends/nxp/backend/ir/lib/tflite/Operator.py:106
↓ 15 callersMethod__init__
(self)
exir/tests/test_memory_planning.py:1125
↓ 15 callersMethod__init__
(self)
test/models/export_program.py:93
↓ 15 callersMethod_adb
(self, cmd, output_callback: Optional[Callable[[str], None]] = None)
backends/qualcomm/export_utils.py:347
↓ 15 callersMethod_emit_node_specific_error
Returns 'err_msg' with node specific information attached.
exir/emit/_emitter.py:259
↓ 15 callersFunction_export
()
devtools/intermediate_output_tap/tests/test_tap_pass.py:36
↓ 15 callersMethod_get_operator
Given a fully qualified name, lookups the operator in the ExecuTorch Program, or adds it if it is not already present
exir/emit/_emitter.py:1477
↓ 15 callersFunction_make_int4_linear
Build an nn.Linear with Int4Tensor weight and return (module, bf16_ref_weight). The bf16 reference is the original unquantized weight, so tests c
backends/cuda/tests/test_int4_dispatch.py:43
↓ 15 callersMethod_prepare_inputs
( batch_size: int = 12, seq_length: int = 7, d_model: int = 768 )
backends/arm/test/models/Gemma3n/test_gemma3nModel.py:144
↓ 15 callersMethod_run_matmul
(self, M, N, K, group_size)
backends/cuda/tests/test_int4_matmul.py:144
↓ 15 callersMethodadd_pre
(self, ops: tflite_model.Operator | list[tflite_model.Operator])
backends/nxp/backend/ir/converter/conversion/common.py:159
↓ 15 callersMethodadd_supported_nodes
(self, node_names)
exir/backend/test/test_group_partitioner.py:42
↓ 15 callersMethodall_versions_for_profile
Return specs for all supported versions of a given profile.
backends/arm/tosa/specification.py:162
↓ 15 callersMethodassert_cmake_cache
( self, key: str, expected_value: str, expected_type: str, )
tools/cmake/common/__init__.py:131
↓ 15 callersMethodcall_operator
(self, op, args, kwargs, meta)
backends/arm/_passes/decompose_log1p_pass.py:65
↓ 15 callersFunctioncheck_exist
(file_or_folder, message=None)
examples/mediatek/aot_utils/llm_utils/sanity_checks.py:70
↓ 15 callersMethodelement_compare
(self, a, b)
devtools/inspector/tests/inspector_test.py:861
↓ 15 callersFunctionexport_and_run_model
( model: nn.Module, example_inputs: Tuple[Any, ...], file_name: str = "CadenceDemoModel", eps_
backends/cadence/aot/export_example.py:101
↓ 15 callersFunctionfilter_fn
(node: torch.fx.Node)
backends/qualcomm/partition/utils.py:36
↓ 15 callersFunctionfn
(shape, dtype)
backends/mlx/test/test_ops.py:4105
↓ 15 callersMethodfrom_bytes
(data: bytes)
backends/vulkan/serialization/vulkan_graph_serialize.py:100
↓ 15 callersMethodget_inputs
(self)
backends/arm/test/ops/test_conv2d.py:113
↓ 15 callersMethodget_num_macs
(self)
backends/nxp/backend/neutron_target_spec.py:138
↓ 15 callersMethodget_output_act_qspec
Get the validated output activation spec. Validate that the output activation qscheme is supported before returning the spec.
backends/arm/quantizer/quantization_config.py:81
↓ 15 callersMethodget_tensor
(self)
exir/emit/test/test_emit.py:1528
↓ 15 callersMethodis_dequant_node
(node: torch.fx.Node)
backends/nxp/neutron_partitioner.py:99
↓ 15 callersFunctionis_distinct
(l)
extension/llm/custom_ops/spinquant/third-party/FFHT/gen.py:10
↓ 15 callersMethodis_valid
Returns true if the extended header appears to be well-formed.
exir/_serialize/_program.py:232
↓ 15 callersMethodlen
(self)
backends/nxp/backend/ir/tflite_generator/meta/meta.py:90
↓ 15 callersFunctionload_dataset
(dataset_path)
examples/qualcomm/oss_scripts/efficientSAM/efficientSAM.py:34
↓ 15 callersFunctionmatch_target
Check if a node's normalized aten target matches the given op. Uses get_aten_target_normalized to handle edge dialect ops. This means sl
backends/mlx/pattern_utils.py:53
↓ 15 callersMethodmin
(self)
devtools/inspector/_inspector.py:303
↓ 15 callersMethodoutput
( self, results: list[ProxyValue], output_kinds: Optional[list[OutputKind]] = None,
backends/test/program_builder.py:79
↓ 15 callersFunctionprint_delegation_info
(graph_module: torch.fx.GraphModule)
devtools/backend_debug/delegation_info.py:179
↓ 15 callersMethodrun_and_verify
(self, model: nn.Module, inputs: Tuple[Any, ...])
examples/cadence/operators/test_g3_ops.py:17
↓ 15 callersMethodset_io
Set quantization_config for input and output nodes. Args: quantization_config (Optional[QuantizationConfig]): Configuration
backends/arm/quantizer/arm_quantizer.py:633
↓ 15 callersFunctiontag_constant_data
Util function for partitioners. This function tags the const/param/buffers nodes whose users all belong within the same partition. This shoul
exir/backend/utils.py:345
↓ 15 callersMethodupdate
(self, *args, **kwargs)
backends/arm/test/tester/test_pipeline.py:88
↓ 15 callersMethodverify_output
( # noqa: C901 self, module: torch.nn.Module, sample_inputs: Tuple[torch.Tensor],
backends/qualcomm/tests/utils.py:347
↓ 14 callersMethodModule
(String moduleAbsolutePath, int loadMode, int numThreads)
extension/android/executorch_android/src/main/java/org/pytorch/executorch/Module.java:58
↓ 14 callersMethodOperatorsLength
(self)
backends/nxp/backend/ir/lib/tflite/SubGraph.py:137
↓ 14 callersMethodValue
(self, j: int)
exir/_serialize/generated/executorch_flatbuffer/CompileSpec.py:41
↓ 14 callersMethod__init__
(self)
exir/backend/test/test_partitioner.py:669
↓ 14 callersMethod__init__
(self, min: List[float] = None)
backends/nxp/backend/ir/tflite_generator/tflite_model.py:183
↓ 14 callersMethod__init__
(self, config)
backends/arm/test/models/Gemma3n/test_gemma3nModel.py:136
↓ 14 callersFunction_ensure_tuple2
(value: Sequence[int])
backends/cortex_m/ops/operators.py:1281
↓ 14 callersMethod_get_op
(self, name, is_edge)
backends/qualcomm/_passes/decompose_atan2.py:51
↓ 14 callersFunction_is_dequantize
(node_: Node)
backends/nxp/backend/edge_helper.py:110
↓ 14 callersFunction_make_qwen3_vl_2b_instruct_layer_config
()
backends/arm/test/models/Qwen3_VL/test_qwen3_vl_layers.py:39
↓ 14 callersFunction_quantize_weights_int4
Quantize [E, N, K] bf16 weight to packed INT4 + scales.
backends/cuda/tests/test_fused_moe.py:60
↓ 14 callersMethod_reset
(self)
extension/llm/custom_ops/test_update_cache.py:35
↓ 14 callersFunction_shape
(node)
backends/qualcomm/_passes/convert_mha_to_sha.py:81
↓ 14 callersFunction_transform
Transforms the program according to the provided passes. Args: self: The ExportedProgram instance to transform *passes: A se
exir/program/_program.py:223
↓ 14 callersMethodactivation
(self)
backends/mediatek/quantizer/qconfig.py:39
↓ 14 callersFunctionannotate_in_out_obs_sharing_op
( node: Node, quantization_config: QuantizationConfig )
backends/qualcomm/quantizer/rules.py:64
↓ 14 callersFunctionannotate_output_qspec
(node: Node, quant_config: QuantizationConfig)
backends/mediatek/quantizer/annotator.py:249
↓ 14 callersFunctionannotate_single_in_share_out
( node: Node, quantization_config: QuantizationConfig )
backends/qualcomm/quantizer/rules.py:93
↓ 14 callersMethodcheckShapeAndDataCapacityConsistency
(int dataCapacity, long[] shape)
extension/android/executorch_android/src/main/java/org/pytorch/executorch/Tensor.java:1052
↓ 14 callersFunctioncreate_node
(graph, target, args, meta, callback=None)
backends/qualcomm/_passes/utils.py:324
↓ 14 callersFunctioncreate_test_pass_manager
Create a test ArmPassManager with a minimal compile spec.
backends/arm/test/passes/test_arm_pass_manager_insertions.py:61
↓ 14 callersFunctioneval_shape
Shape maybe immutable so we return a new shape. Return None for dimensions that are unbacked e.g. first dimension of nonzero's output.
exir/sym_util.py:73
↓ 14 callersFunctionexpand_around_channel
Expand a scalar or 1-D parameter around the channel dimension into a broadcastable shape while preserving the channel location.
backends/arm/_passes/arm_pass_utils.py:128
↓ 14 callersFunctionexport_to_edge
( model: Union[torch.fx.GraphModule, torch.nn.Module], example_inputs: Tuple[Value, ...], *, e
extension/export_util/utils.py:74
↓ 14 callersFunctiongenerate_qnn_executorch_option
( compiler_specs: List[CompileSpec], )
backends/qualcomm/partition/utils.py:16
↓ 14 callersMethodget_example_inputs
(self)
backends/cortex_m/test/tester.py:156
↓ 14 callersFunctioninternal_assert
This is exir's custom assert method. It internally just throws InternalError. Note that the sole purpose is to throw our own error while main
exir/error.py:24
↓ 14 callersFunctionmake_simple_generated_case
(*args, torch_fn)
kernels/test/test_case_gen.py:282
↓ 14 callersMethodrun
Executes this stage, generates the 'artifact', for later stages.
backends/test/harness/stages/stage.py:35
↓ 14 callersMethodsetup_cache
Setup key value caches for attention calculation. If called after kv_cache is already setup, this will be skipped. Args:
extension/llm/modules/attention.py:165
↓ 14 callersFunctionstride_from_dim_order
Converts dim order to stride using sizes e.g. if sizes = (2, 3, 4) and dim_order = (0, 1, 2) then strides = (12, 4, 1) while for the same
exir/tensor.py:104
↓ 14 callersMethodto_tensor
(self)
exir/pass_base.py:141
↓ 14 callersMethodunion
(self, x, y)
exir/backend/canonical_partitioners/config_partitioner.py:125
↓ 13 callersMethod_edb
(self, cmd)
backends/samsung/test/utils/runtime_executor.py:47
↓ 13 callersFunction_export_graph_module
()
backends/cortex_m/test/misc/test_node_finders.py:23
↓ 13 callersFunction_has_op
(gm, target)
backends/mlx/test/test_passes.py:71
↓ 13 callersFunction_is_quantize
(node_: Node)
backends/nxp/backend/edge_helper.py:119
↓ 13 callersMethod_make_args
(self, n_layers: int = 4, num_kv_shared_layers: int = 0)
examples/models/llama/tests/test_yoco_attention.py:351
↓ 13 callersFunction_make_fake_tensor
(shape, dtype=torch.float32)
backends/vulkan/test/test_vulkan_tensor_repr.py:43
↓ 13 callersFunction_prepare_for_quantization
(exported_model, is_qat: bool = False)
backends/nxp/tests/generic_tests/test_quantizer.py:90
↓ 13 callersFunction_program_flatbuffer_to_json
Converts binary flatbuffer data into Program-compatible JSON. The binary is parsed using the schema in //executorch/schema/program.fbs.
exir/_serialize/_flatbuffer.py:466
↓ 13 callersMethodadd_absolute_placement_constraint
Add a memory pinning constraint for `node` to `mem_id`.
backends/cadence/aot/memory_constraints.py:307
↓ 13 callersMethodcalculate_cache_key
(layer_id: int, head_id: int)
examples/models/llama/static_attention.py:57
↓ 13 callersFunctioncomposite_step
( buf_name, log_n, from_it, to_it, log_w, registers, move_instruction, special
extension/llm/custom_ops/spinquant/third-party/FFHT/gen.py:415
↓ 13 callersMethodcpp_type
(self, *, strip_ref: bool = False)
codegen/api/types/types.py:60
↓ 13 callersFunctionformat_delegated_graph
Return the formatted graph string of including lowered_module (both backend id and original graph) together with the graph module. Example output
exir/backend/utils.py:293
↓ 13 callersMethodget_arg_repset
(self, i: int)
backends/vulkan/utils.py:1489
↓ 13 callersMethodget_inputs
(self)
backends/arm/test/ops/test_conv3d.py:114
← previousnext →401–500 of 28,493, ranked by callers