MCPcopy Create free account

hub / github.com/pytorch/executorch / functions

Functions28,493 in github.com/pytorch/executorch

↓ 11 callersMethodto_dict
(self)
backends/mlx/pte_inspector.py:75
↓ 11 callersMethodtry_constrain_with_arg_repset
Attempt to constrain the repsets of the tensors participating in this operator based on an "existing" repset of an argument. The exis
backends/vulkan/utils.py:1495
↓ 11 callersFunctionunpack_gguf_tensor
Unpack a single GGUF tensor. Returns an ``Int4Tensor`` for Q4_K, ``IntxUnpackedToInt8Tensor`` for Q6_K, or a plain ``torch.Tensor`` for F32/F
examples/models/gemma4_31b/quant/gguf.py:165
↓ 10 callersMethod__init__
(self)
exir/program/test/test_program.py:61
↓ 10 callersMethod__init__
( self, exported_program: torch.export.ExportedProgram, buf_data: torch.Te
backends/transforms/test/test_create_mutable_buffer.py:103
↓ 10 callersMethod__init__
(self, config: GemmaConfig)
examples/mediatek/models/llm_models/modeling_gemma.py:63
↓ 10 callersFunction_count_rescales
(graph_module: torch.fx.GraphModule)
backends/arm/test/passes/test_rescale_optimization.py:89
↓ 10 callersFunction_expr_equals
Return True if the SymPy expressions are equivalent.
backends/arm/test/misc/tosa_dialect/test_tosa_shape_ops.py:37
↓ 10 callersMethod_get_intermediate_path
Gets the path used for dumping intermediate results such as tosa and pte. Returns: Path where intermediate results are sa
backends/arm/common/arm_compile_spec.py:298
↓ 10 callersFunction_get_quantization_annotation
(node: Node)
backends/samsung/quantizer/annotator.py:231
↓ 10 callersMethod_get_test_model
(self)
backends/apple/coreml/test/test_torch_ops.py:36
↓ 10 callersFunction_is_pure_dtype_cast
Check that _to_copy kwargs only specify dtype (no device/layout/memory_format).
backends/mlx/passes.py:375
↓ 10 callersMethod_make_unary_op
Create an OpRepSets for a simple unary op (single tensor in, single tensor out).
backends/vulkan/test/test_vulkan_tensor_repr.py:562
↓ 10 callersMethod_test_sdpa_common
( self, n_heads_kv, n_heads_q, head_dim, max_seq_len, seq_len,
extension/llm/custom_ops/test_sdpa_with_kv_cache.py:446
↓ 10 callersMethod_validate
Validate the configuration against VGF-supported TOSA profiles.
backends/arm/vgf/compile_spec.py:45
↓ 10 callersFunction_write_yaml
(path: Path, payload: dict)
codegen/tools/test/test_gen_max_kernel_num.py:37
↓ 10 callersMethodadd
(self, ngram: Tuple[int])
examples/qualcomm/oss_scripts/llama/decoder_utils.py:148
↓ 10 callersMethodadd_exported_program
Add exported program to the ETRecord after it has been created. This method allows users to add an exported program they want to rec
devtools/etrecord/_etrecord.py:401
↓ 10 callersFunctioncalibrate_and_quantize
Quantize the provided model. :param model: Aten model (or it's GraphModule representation) to quantize. :param calibration_inputs: An iterato
backends/nxp/quantizer/utils.py:206
↓ 10 callersFunctioncheck_ext
(file, ext, message=None)
examples/mediatek/aot_utils/llm_utils/sanity_checks.py:78
↓ 10 callersFunctionconvert_to_float_tensor
Convert input_data into a torch.Tensor on CPU with dtype torch.float64. This function handles the following types of input: - Scalar (int
devtools/inspector/_inspector_utils.py:988
↓ 10 callersMethodcreate_transpose_operator_after
Create a TFLite 'Transpose' operator after the 'after_operator'. The output of 'after_operator' at index 'on_output_index' is
backends/nxp/backend/ir/converter/builder/model_builder.py:984
↓ 10 callersMethoddecode
Decode.
examples/mediatek/aot_utils/llm_utils/tokenizers_/tokenization_qwen2.py:294
↓ 10 callersMethoddecoder
`tokenizers.decoders.Decoder`: The Rust decoder for this tokenizer.
examples/mediatek/aot_utils/llm_utils/tokenizers_/tokenization_utils_fast.py:282
↓ 10 callersMethoddump_executorch_program
Prints the ExecuTorch binary in a human readable format. Args: verbose (bool): If False prints the binar
exir/program/_program.py:1938
↓ 10 callersFunctionemit_lifted_constant
Lift a scalar to a 0-D tensor. Concrete scalars (int/float/bool) become deduplicated constants. Dynamic values (SymInt Slots) emit a FullNode
backends/mlx/builder/op_helpers.py:150
↓ 10 callersFunctionemit_program
Given a exported program, it returns the program in the format of the Python version of the flatbuffer Program schema. Args: met
exir/emit/_emit_program.py:118
↓ 10 callersFunctionevaluate_symbolic_expr_values
Return a best-effort finite set of possible integer values. The helper first relies on ``bound_sympy`` for cheap singleton detection. When in
backends/arm/_passes/symbolic_value_range.py:101
↓ 10 callersMethodfind_pattern_matches
Match all given patterns in the graph and return match results with acceptance/rejection status. Each node can only be part of one match,
backends/cortex_m/quantizer/pattern_matcher.py:172
↓ 10 callersFunctionfor_args
(f, a)
exir/serde/export_serialize.py:2467
↓ 10 callersMethodfromBlob
Creates a new Tensor instance with dtype torch.int8 with specified shape and data as array of bytes. @param data Tensor elements @param shape Tensor
extension/android/executorch_android/src/main/java/org/pytorch/executorch/Tensor.java:160
↓ 10 callersFunctionget_default_passes
Returns a list of passes that are enabled by default for the MLX backend.
backends/mlx/passes.py:28
↓ 10 callersFunctionget_dim_order
Given a memory_format and a tensor rank, generate a dim_order
exir/dim_order_utils.py:52
↓ 10 callersFunctionget_passes_dependency_for_capture_program
This function records the dependencies for passes used in the to_edge_transform_and_lower_to_qnn. It returns a dictionary where the keys are
backends/qualcomm/_passes/utils.py:51
↓ 10 callersFunctionget_random_calibration_inputs
( input_spec: Iterable[ModelInputSpec], num_samples: int = 4 )
backends/nxp/tests/executorch_pipeline.py:87
↓ 10 callersMethodhas_overlap
r""" The passed in intervals are inclusive in both sides. Return if they have overlapping.
exir/memory_planning.py:93
↓ 10 callersMethodhas_same_packed_dim_info_set
Check if self and other produce the exact same sets of PackedDimInfo for both buffer and texture storage types. Completely empty reps
backends/vulkan/utils.py:1025
↓ 10 callersMethodis_constrained
A "constrained" RepSet is one that has either: 1. A single valid texture memory layout, and no valid buffer memory layouts 2.
backends/vulkan/utils.py:987
↓ 10 callersMethodis_memory_planned
Return true if the node is either (1) a parameter, or (2) a placeholder.
backends/cadence/aot/memory_constraints.py:164
↓ 10 callersMethodis_nhwc_node
(node: torch.fx.Node)
backends/xnnpack/_passes/channels_last_tagged_reshape_pass.py:109
↓ 10 callersMethodlog_skip
(self, node: torch.fx.Node, reason: str)
backends/vulkan/partitioner/vulkan_partitioner.py:192
↓ 10 callersMethodlower_and_run_model
( self, model: torch.nn.Module, inputs: Any, generate_random_test_inputs=True,
backends/test/suite/conftest.py:73
↓ 10 callersMethodmake_intersect
Merge this TensorRepr with another TensorRepr, returning a new TensorRepr with the intersection of the two.
backends/vulkan/utils.py:911
↓ 10 callersFunctionmake_module_wrapper
( name: str, module_factory: Callable[[], torch.nn.Module] )
backends/arm/test/models/test_nn_modules.py:30
↓ 10 callersFunctionop_node_count
(graph_module: torch.fx.GraphModule, canonical_op_name: str)
backends/vulkan/test/test_vulkan_passes.py:80
↓ 10 callersMethodpartition
(self, partition_stage: Optional[Partition] = None)
backends/test/harness/tester.py:246
↓ 10 callersFunctionquantize_module
( module: torch.nn.Module, inputs, calibration_dataset, precision: Precision, is_per_chann
backends/samsung/utils/export_utils.py:55
↓ 10 callersFunctionquantize_multiplier_aot
(scale: float)
backends/cortex_m/passes/passes_utils.py:193
↓ 10 callersFunctionreinplace_pass
Rewrite functional ops in-place when safe. Walks the graph in reverse topological order. For each `call_function` node whose target is in the
exir/passes/reinplace.py:269
↓ 10 callersFunctionserialize
( exported_program: ep.ExportedProgram, opset_version: Optional[Dict[str, int]] = None, )
exir/serde/serialize.py:701
↓ 10 callersMethodserialize_input
( self, arg, arg_type: Optional[torch._C.Argument] = None )
exir/serde/export_serialize.py:698
↓ 10 callersFunctionstrip_taps_
Replace every `tap.Tensor(src, reducer_name)` node in every method of `edge_manager` with the materialised reducer subgraph, in place. A
devtools/intermediate_output_tap/_strip_pass.py:26
↓ 10 callersMethodto_bytes
(self)
backends/vulkan/serialization/vulkan_graph_serialize.py:150
↓ 10 callersFunctiontorch_type_to_numpy_type
Convert Torch data type to Numpy data type.
backends/nxp/backend/ir/converter/conversion/translator.py:550
↓ 10 callersFunctiontransform_and_check_numerics
Run a graph transformation and validate numerical equivalence. Args: original_graph: The original graph module before transformation
backends/cadence/aot/tests/test_reorder_ops_passes.py:38
↓ 10 callersFunctiontry_get_tensor_constant_from_node
Get the static data from a given node. If it doesn't have any data, return `None`.
backends/nxp/backend/edge_helper.py:94
↓ 10 callersMethodupdate
( self, input_pos: torch.Tensor, k_val: torch.Tensor, v_val: torch.Tensor )
examples/models/llama/attention.py:301
↓ 10 callersFunctionupdate_spill_fill_size
( exported_program: ExportedProgram | List[LoweredBackendModule], )
backends/qualcomm/utils/utils.py:229
↓ 10 callersMethodwrap
(arg: Value, i: int)
exir/tracer.py:509
↓ 9 callersMethodBuffer
(self)
backends/nxp/backend/ir/lib/tflite/Tensor.py:74
↓ 9 callersMethod__init__
(self, fn)
exir/tests/test_verification.py:23
↓ 9 callersMethod__init__
(self)
backends/cortex_m/test/models/test_nn_modules.py:38
↓ 9 callersMethod__init__
( self, module: torch.nn.Module, test_data: T, aten_op: str | List[str],
backends/arm/test/tester/test_pipeline.py:1162
↓ 9 callersMethod_adjust_pad_if_needed
Adjust padding to satisfy TOSA's integer output-size requirement. Torch ``Conv2d`` does not require the result of ``(input + 2 * pad
backends/arm/_passes/rewrite_conv_pass.py:60
↓ 9 callersMethod_compare_eager_quantized_model_outputs
Utility to compare eager quantized model output with session output after coreml lowering
backends/apple/coreml/test/test_coreml_recipes.py:50
↓ 9 callersMethod_compare_eager_unquantized_model_outputs
Utility to compare eager unquantized model output with session output using SQNR
backends/apple/coreml/test/test_coreml_recipes.py:61
↓ 9 callersFunction_dummy_qconfig
()
backends/cortex_m/test/misc/test_pattern_matcher.py:62
↓ 9 callersFunction_flatc_compile
Serializes JSON data to a binary flatbuffer file. Args: output_dir: Directory under which to create the binary flatbuffer file. s
exir/_serialize/_flatbuffer.py:333
↓ 9 callersMethod_get_default_stage
(self, stage_type: StageType, *args, **kwargs)
backends/test/harness/tester.py:112
↓ 9 callersFunction_is_call
(node)
backends/qualcomm/_passes/convert_mha_to_sha.py:42
↓ 9 callersFunction_json_to_program
Returns a Program deserialized from the given JSON string.
exir/_serialize/_program.py:89
↓ 9 callersFunction_make_op_node
Create a mock torch.fx.Node for use in OpRepSets tests.
backends/vulkan/test/test_vulkan_tensor_repr.py:48
↓ 9 callersMethod_prepare_model
Prepare ToyModelForMemPlanning through SpecPropPass + ToOutVarPass.
exir/tests/test_memory_planning.py:1430
↓ 9 callersFunction_progress_newline
End a progress line.
backends/qualcomm/scripts/download_qnn_sdk.py:41
↓ 9 callersMethod_run_dequant
(self, N, K, group_size)
backends/cuda/tests/test_int4_matmul.py:99
↓ 9 callersMethod_run_stage
(self, stage_instance, inputs=None, *args, **kwargs)
backends/test/harness/tester.py:201
↓ 9 callersMethod_test
( self, test_set: Dict[str, Tuple[ExportedProgram, Partitioner, List[str]]] )
exir/backend/test/test_to_backend_multi_method.py:94
↓ 9 callersMethodadd
(self, x)
examples/models/llama/static_attention.py:242
↓ 9 callersMethodadd_executorch_program
Add executorch program data to the ETRecord after it has been created. This method allows users to add executorch program data they
devtools/etrecord/_etrecord.py:349
↓ 9 callersMethodargumenttype_evalue_convert
Takes in the type, name and mutability corresponding to an argument, and generates a tuple of: (1) the C++ code necessary to unbox th
codegen/api/unboxing.py:80
↓ 9 callersMethodavg
(self)
devtools/inspector/_inspector.py:299
↓ 9 callersFunctionbuild_random_tiny_model
()
examples/models/gemma4_31b/tests/test_pipeline.py:129
↓ 9 callersMethodcollect_executorch_and_eager_outputs
Compares the output of the given eager mode PyTorch model with the output of the equivalent executorch model, both provided with exam
examples/models/test/test_export.py:21
↓ 9 callersMethodcopy_with_new_data
(self, new_data: Any)
export/stages.py:48
↓ 9 callersMethoddecode
( self, codes, k_cache, v_cache, end_index, end_offset,
examples/qualcomm/oss_scripts/moshi/model/static_mimi.py:566
↓ 9 callersMethoddefault_stage_classes
Returns a map of StageType to default Stage implementation.
backends/test/harness/tester.py:97
↓ 9 callersMethoddeserialize_input
(self, inp: Argument)
exir/serde/export_serialize.py:1943
↓ 9 callersMethoddump_debug_info
Dump debugging information into the intermediates path. Args: debug_mode: The debug mode to use for dumping debug information.
backends/arm/common/arm_compile_spec.py:319
↓ 9 callersMethodfrom_bytes
Tries to read an extended header from the provided data. Does not validate that the header is well-formed. Callers should use is_vali
exir/_serialize/_program.py:195
↓ 9 callersFunctionfrom_context_binary
( # noqa: C901 ctx_path: str | bytes, op_name: str, soc_model: QcomChipset = QcomChipset.SM8650,
backends/qualcomm/utils/utils.py:794
↓ 9 callersMethodfrom_q_dq_node
( cls, quant_node: torch.fx.Node, ep: Optional[ExportedProgram] = None )
backends/xnnpack/operators/quant_params.py:235
↓ 9 callersMethodget
(self, var: str)
tools/cmake/cmake_cache.py:27
↓ 9 callersFunctionget_args
()
backends/arm/test/test_model.py:14
↓ 9 callersFunctionget_config_methods
()
exir/program/test/test_program.py:177
↓ 9 callersMethodget_executorch_program_manager
Get the ExecutorchProgramManager. Returns: The ExecutorchProgramManager Raises: RuntimeError: If th
export/export.py:550
↓ 9 callersFunctionget_exported_programs
()
exir/program/test/test_program.py:150
↓ 9 callersMethodget_inputs
(self)
backends/arm/test/ops/test_tril.py:41
↓ 9 callersMethodget_inputs
(self)
backends/arm/test/ops/test_conv1d.py:109
↓ 9 callersFunctionget_node_tensor_specs
r""" Return the list of the tensor specs for the node or empty list if the node has no tensor specs.
exir/memory_planning.py:892
↓ 9 callersMethodget_operators
(self)
backends/nxp/backend/ir/converter/builder/model_builder.py:1701
← previousnext →601–700 of 28,493, ranked by callers