mirror of
https://github.com/pytorch/pytorch.git
synced 2025-10-21 05:34:18 +08:00
2460dced8f
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/28944 Add torch.nn.GELU for GELU activation Test Plan: buck test mode/dev-nosan //caffe2/test:nn -- "GELU" Reviewed By: hl475, houseroad Differential Revision: D18240946 fbshipit-source-id: 6284b30def9bd4c12bf7fb2ed08b1b2f0310bb78
194 lines
8.3 KiB
Python
194 lines
8.3 KiB
Python
import torch
|
|
|
|
from cpp_api_parity import TorchNNModuleMetadata
|
|
|
|
# NOTE: In order to let Python/C++ API parity test pass for any of the modules here,
|
|
# after fixing the C++ module implementation you should change change its "Implementation Parity"
|
|
# from "No" to "Yes" in parity-tracker.md, and change the module's `TorchNNModuleMetadata` here
|
|
# to look like:
|
|
#
|
|
# 'Linear': TorchNNModuleMetadata(
|
|
# cpp_default_constructor_args="(3, 4)",
|
|
# num_attrs_recursive=4,
|
|
# )
|
|
#
|
|
# The meaning of each field of `TorchNNModuleMetadata` is as follows:
|
|
#
|
|
# `cpp_default_constructor_args`: string that represents the required non-keyword arguments
|
|
# for the C++ module constructor. For example, since `LinearOptions` expects two non-keyword
|
|
# arguments `(in_features, out_features)`, the `cpp_default_constructor_args` for `Linear`
|
|
# will be the string representation of any integer 2-tuple, such as "(3, 4)".
|
|
# Note that the C++ module constructor must take the exact same number of non-keyword arguments
|
|
# as the Python module constructor.
|
|
#
|
|
# `num_attrs_recursive`: the number of attributes (including parameters, buffers and non-tensor
|
|
# attributes, but excluding the attributes in `python_ignored_attrs`) of the Python module.
|
|
# If the module contains any submodule, the submodule's attributes also need to be counted.
|
|
#
|
|
# `python_ignored_constructor_args`: (optional) list of Python constructor args that are
|
|
# ignored in Python/C++ API parity test.
|
|
#
|
|
# `python_ignored_attrs`: (optional) list of Python module attributes (including parameters,
|
|
# buffers and non-tensor attributes) that are ignored in Python/C++ API parity test.
|
|
#
|
|
# `python_optional_attribute_to_jit_type`: (optional) map between Python None-able module
|
|
# attribute to its corresponding JIT type. For example, in `AvgPool2d`:
|
|
# { "divisor_override": torch._C.OptionalType(torch._C.IntType.get()) }
|
|
module_metadata_map = {
|
|
'Conv1d': TorchNNModuleMetadata(),
|
|
'Conv2d': TorchNNModuleMetadata(),
|
|
'Conv3d': TorchNNModuleMetadata(),
|
|
'ConvTranspose1d': TorchNNModuleMetadata(),
|
|
'ConvTranspose2d': TorchNNModuleMetadata(),
|
|
'ConvTranspose3d': TorchNNModuleMetadata(),
|
|
'Unfold': TorchNNModuleMetadata(),
|
|
'Fold': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="(3, 2)",
|
|
num_attrs_recursive=5,
|
|
),
|
|
'MaxPool1d': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="(2)",
|
|
num_attrs_recursive=5,
|
|
python_ignored_constructor_args=['return_indices'],
|
|
python_ignored_attrs=['return_indices'],
|
|
),
|
|
'MaxPool2d': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="(2)",
|
|
num_attrs_recursive=5,
|
|
python_ignored_constructor_args=['return_indices'],
|
|
python_ignored_attrs=['return_indices'],
|
|
),
|
|
'MaxPool3d': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="(2)",
|
|
num_attrs_recursive=5,
|
|
python_ignored_constructor_args=['return_indices'],
|
|
python_ignored_attrs=['return_indices'],
|
|
),
|
|
'MaxUnpool1d': TorchNNModuleMetadata(),
|
|
'MaxUnpool2d': TorchNNModuleMetadata(),
|
|
'MaxUnpool3d': TorchNNModuleMetadata(),
|
|
'AvgPool1d': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="(2)",
|
|
num_attrs_recursive=5,
|
|
),
|
|
'AvgPool2d': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="(2)",
|
|
num_attrs_recursive=6,
|
|
python_optional_attribute_to_jit_type={
|
|
"divisor_override": torch._C.OptionalType(torch._C.IntType.get()),
|
|
}
|
|
),
|
|
'AvgPool3d': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="(2)",
|
|
num_attrs_recursive=6,
|
|
python_optional_attribute_to_jit_type={
|
|
"divisor_override": torch._C.OptionalType(torch._C.IntType.get()),
|
|
}
|
|
),
|
|
'FractionalMaxPool2d': TorchNNModuleMetadata(),
|
|
'LPPool1d': TorchNNModuleMetadata(),
|
|
'LPPool2d': TorchNNModuleMetadata(),
|
|
'AdaptiveMaxPool1d': TorchNNModuleMetadata(),
|
|
'AdaptiveMaxPool2d': TorchNNModuleMetadata(),
|
|
'AdaptiveMaxPool3d': TorchNNModuleMetadata(),
|
|
'AdaptiveAvgPool1d': TorchNNModuleMetadata(),
|
|
'AdaptiveAvgPool2d': TorchNNModuleMetadata(),
|
|
'AdaptiveAvgPool3d': TorchNNModuleMetadata(),
|
|
'ReflectionPad1d': TorchNNModuleMetadata(),
|
|
'ReflectionPad2d': TorchNNModuleMetadata(),
|
|
'ReplicationPad1d': TorchNNModuleMetadata(),
|
|
'ReplicationPad2d': TorchNNModuleMetadata(),
|
|
'ReplicationPad3d': TorchNNModuleMetadata(),
|
|
'ZeroPad2d': TorchNNModuleMetadata(),
|
|
'ConstantPad1d': TorchNNModuleMetadata(),
|
|
'ConstantPad2d': TorchNNModuleMetadata(),
|
|
'ConstantPad3d': TorchNNModuleMetadata(),
|
|
'ELU': TorchNNModuleMetadata(),
|
|
'Hardshrink': TorchNNModuleMetadata(),
|
|
'Hardtanh': TorchNNModuleMetadata(),
|
|
'LeakyReLU': TorchNNModuleMetadata(),
|
|
'LogSigmoid': TorchNNModuleMetadata(),
|
|
'MultiheadAttention': TorchNNModuleMetadata(),
|
|
'PReLU': TorchNNModuleMetadata(),
|
|
'ReLU': TorchNNModuleMetadata(),
|
|
'ReLU6': TorchNNModuleMetadata(),
|
|
'RReLU': TorchNNModuleMetadata(),
|
|
'SELU': TorchNNModuleMetadata(),
|
|
'CELU': TorchNNModuleMetadata(),
|
|
'GELU': TorchNNModuleMetadata(),
|
|
'Sigmoid': TorchNNModuleMetadata(),
|
|
'Softplus': TorchNNModuleMetadata(),
|
|
'Softshrink': TorchNNModuleMetadata(),
|
|
'Softsign': TorchNNModuleMetadata(),
|
|
'Tanh': TorchNNModuleMetadata(),
|
|
'Tanhshrink': TorchNNModuleMetadata(),
|
|
'Threshold': TorchNNModuleMetadata(),
|
|
'Softmin': TorchNNModuleMetadata(),
|
|
'Softmax': TorchNNModuleMetadata(),
|
|
'Softmax2d': TorchNNModuleMetadata(),
|
|
'LogSoftmax': TorchNNModuleMetadata(),
|
|
'AdaptiveLogSoftmaxWithLoss': TorchNNModuleMetadata(),
|
|
'BatchNorm1d': TorchNNModuleMetadata(),
|
|
'BatchNorm2d': TorchNNModuleMetadata(),
|
|
'BatchNorm3d': TorchNNModuleMetadata(),
|
|
'GroupNorm': TorchNNModuleMetadata(),
|
|
'SyncBatchNorm': TorchNNModuleMetadata(),
|
|
'InstanceNorm1d': TorchNNModuleMetadata(),
|
|
'InstanceNorm2d': TorchNNModuleMetadata(),
|
|
'InstanceNorm3d': TorchNNModuleMetadata(),
|
|
'LayerNorm': TorchNNModuleMetadata(),
|
|
'LocalResponseNorm': TorchNNModuleMetadata(),
|
|
'RNN': TorchNNModuleMetadata(),
|
|
'LSTM': TorchNNModuleMetadata(),
|
|
'GRU': TorchNNModuleMetadata(),
|
|
'RNNCell': TorchNNModuleMetadata(),
|
|
'LSTMCell': TorchNNModuleMetadata(),
|
|
'GRUCell': TorchNNModuleMetadata(),
|
|
'Transformer': TorchNNModuleMetadata(),
|
|
'TransformerEncoder': TorchNNModuleMetadata(),
|
|
'TransformerDecoder': TorchNNModuleMetadata(),
|
|
'TransformerEncoderLayer': TorchNNModuleMetadata(),
|
|
'TransformerDecoderLayer': TorchNNModuleMetadata(),
|
|
'Identity': TorchNNModuleMetadata(),
|
|
'Linear': TorchNNModuleMetadata(),
|
|
'Bilinear': TorchNNModuleMetadata(),
|
|
'Dropout': TorchNNModuleMetadata(),
|
|
'Dropout2d': TorchNNModuleMetadata(),
|
|
'Dropout3d': TorchNNModuleMetadata(),
|
|
'AlphaDropout': TorchNNModuleMetadata(),
|
|
'Embedding': TorchNNModuleMetadata(),
|
|
'EmbeddingBag': TorchNNModuleMetadata(),
|
|
'CosineSimilarity': TorchNNModuleMetadata(),
|
|
'PairwiseDistance': TorchNNModuleMetadata(),
|
|
'L1Loss': TorchNNModuleMetadata(
|
|
cpp_default_constructor_args="()",
|
|
num_attrs_recursive=1,
|
|
python_ignored_constructor_args=['size_average', 'reduce'],
|
|
),
|
|
'MSELoss': TorchNNModuleMetadata(),
|
|
'CrossEntropyLoss': TorchNNModuleMetadata(),
|
|
'CTCLoss': TorchNNModuleMetadata(),
|
|
'NLLLoss': TorchNNModuleMetadata(),
|
|
'PoissonNLLLoss': TorchNNModuleMetadata(),
|
|
'KLDivLoss': TorchNNModuleMetadata(),
|
|
'BCELoss': TorchNNModuleMetadata(),
|
|
'BCEWithLogitsLoss': TorchNNModuleMetadata(),
|
|
'MarginRankingLoss': TorchNNModuleMetadata(),
|
|
'HingeEmbeddingLoss': TorchNNModuleMetadata(),
|
|
'MultiLabelMarginLoss': TorchNNModuleMetadata(),
|
|
'SmoothL1Loss': TorchNNModuleMetadata(),
|
|
'SoftMarginLoss': TorchNNModuleMetadata(),
|
|
'MultiLabelSoftMarginLoss': TorchNNModuleMetadata(),
|
|
'CosineEmbeddingLoss': TorchNNModuleMetadata(),
|
|
'MultiMarginLoss': TorchNNModuleMetadata(),
|
|
'TripletMarginLoss': TorchNNModuleMetadata(),
|
|
'PixelShuffle': TorchNNModuleMetadata(),
|
|
'Upsample': TorchNNModuleMetadata(),
|
|
'UpsamplingNearest2d': TorchNNModuleMetadata(),
|
|
'UpsamplingBilinear2d': TorchNNModuleMetadata(),
|
|
'Flatten': TorchNNModuleMetadata(),
|
|
'CrossMapLRN2d': TorchNNModuleMetadata(),
|
|
'FractionalMaxPool3d': TorchNNModuleMetadata(),
|
|
'GLU': TorchNNModuleMetadata(),
|
|
}
|