-
Notifications
You must be signed in to change notification settings - Fork 429
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* add python, onnx model, burn/node for conv_transpose1d * modify supported-onnx-ops.md about convtranspose1d * add conv_transpose1d function in op_configuration and to_burn * add test and modify dim_inference * apply cargo fmt to formatting * fix reviewer point-outs by adding symetry checks for padding in 2d,3d * fix pads initilization and check ways
- Loading branch information
Showing
11 changed files
with
425 additions
and
15 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Binary file added
BIN
+572 Bytes
crates/burn-import/onnx-tests/tests/conv_transpose1d/conv_transpose1d.onnx
Binary file not shown.
57 changes: 57 additions & 0 deletions
57
crates/burn-import/onnx-tests/tests/conv_transpose1d/conv_transpose1d.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,57 @@ | ||
| #!/usr/bin/env python3 | ||
|
|
||
| # used to generate model: conv_transpose1d.onnx | ||
|
|
||
| import torch | ||
| import torch.nn as nn | ||
|
|
||
|
|
||
| class Model(nn.Module): | ||
| def __init__(self): | ||
| super(Model, self).__init__() | ||
| self.transposed_conv = nn.ConvTranspose1d( | ||
| in_channels=4, | ||
| out_channels=6, | ||
| kernel_size=3, | ||
| stride=2, | ||
| padding=1, | ||
| dilation=2, | ||
| output_padding=1, | ||
| groups=2 | ||
| ) | ||
|
|
||
| def forward(self, x): | ||
| return self.transposed_conv(x) | ||
|
|
||
|
|
||
| def main(): | ||
|
|
||
| # Set seed for reproducibility | ||
| torch.manual_seed(42) | ||
|
|
||
| torch.set_printoptions(precision=8) | ||
|
|
||
| # Export to onnx | ||
| model = Model() | ||
| model.eval() | ||
| device = torch.device("cpu") | ||
|
|
||
| file_name = "conv_transpose1d.onnx" | ||
| test_input = torch.ones(2, 4, 10, device=device) | ||
| torch.onnx.export(model, test_input, file_name, | ||
| verbose=False, opset_version=16) | ||
|
|
||
| print("Finished exporting model to {}".format(file_name)) | ||
|
|
||
| # Output some test data for use in the test | ||
| print("Test input data shape of ones: {}".format(test_input.shape)) | ||
| output = model.forward(test_input) | ||
| print("Test output data shape: {}".format(output.shape)) | ||
|
|
||
| sum = output.sum().item() | ||
|
|
||
| print("Test output sum: {}".format(sum)) | ||
|
|
||
|
|
||
| if __name__ == "__main__": | ||
| main() |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,201 @@ | ||
| use super::{Node, NodeCodegen, SerializationBackend}; | ||
| use crate::burn::{BurnImports, OtherType, Scope, TensorType, ToTokens, Type}; | ||
| use burn::{ | ||
| module::{ConstantRecord, Param, ParamId}, | ||
| nn::conv::{ConvTranspose1dConfig, ConvTranspose1dRecord}, | ||
| record::{PrecisionSettings, Record}, | ||
| tensor::{Tensor, TensorData}, | ||
| }; | ||
| use proc_macro2::TokenStream; | ||
| use quote::quote; | ||
| use serde::Serialize; | ||
|
|
||
| #[derive(Debug, Clone)] | ||
| pub struct ConvTranspose1dNode { | ||
| pub field: OtherType, | ||
| pub input: TensorType, | ||
| pub output: TensorType, | ||
| pub data_weights: TensorData, | ||
| pub data_bias: Option<TensorData>, | ||
| pub config: ConvTranspose1dConfig, | ||
| } | ||
|
|
||
| impl ConvTranspose1dNode { | ||
| pub fn new<S: AsRef<str>>( | ||
| name: S, | ||
| input: TensorType, | ||
| output: TensorType, | ||
| data_weights: TensorData, | ||
| data_bias: Option<TensorData>, | ||
| config: ConvTranspose1dConfig, | ||
| ) -> Self { | ||
| Self { | ||
| field: OtherType::new( | ||
| name, | ||
| quote! { | ||
| ConvTranspose1d<B> | ||
| }, | ||
| ), | ||
| input, | ||
| output, | ||
| data_weights, | ||
| data_bias, | ||
| config, | ||
| } | ||
| } | ||
| } | ||
|
|
||
| impl<PS: PrecisionSettings> NodeCodegen<PS> for ConvTranspose1dNode { | ||
| fn input_types(&self) -> Vec<Type> { | ||
| vec![Type::Tensor(self.input.clone())] | ||
| } | ||
| fn output_types(&self) -> Vec<Type> { | ||
| vec![Type::Tensor(self.output.clone())] | ||
| } | ||
| fn field_type(&self) -> Option<Type> { | ||
| Some(Type::Other(self.field.clone())) | ||
| } | ||
|
|
||
| fn field_init(&self) -> Option<TokenStream> { | ||
| let name = &self.field.name; | ||
| let channels = self.config.channels.to_tokens(); | ||
| let kernel_size = self.config.kernel_size.to_tokens(); | ||
| let stride = self.config.stride.to_tokens(); | ||
| let dilation = self.config.dilation.to_tokens(); | ||
| let groups = self.config.groups.to_tokens(); | ||
| let padding = self.config.padding.to_tokens(); | ||
| let padding_out = self.config.padding_out.to_tokens(); | ||
| let bias = self.config.bias; | ||
|
|
||
| let tokens = quote! { | ||
| let #name = ConvTranspose1dConfig::new(#channels, #kernel_size) | ||
| .with_stride(#stride) | ||
| .with_padding(#padding) | ||
| .with_padding_out(#padding_out) | ||
| .with_dilation(#dilation) | ||
| .with_groups(#groups) | ||
| .with_bias(#bias) | ||
| .init(device); | ||
| }; | ||
|
|
||
| Some(tokens) | ||
| } | ||
|
|
||
| fn field_serialize<S: serde::Serializer>(&self, serializer: S) -> Result<S::Ok, S::Error> { | ||
| let device = Default::default(); | ||
| let record = ConvTranspose1dRecord::<SerializationBackend> { | ||
| weight: Param::initialized( | ||
| ParamId::new(), | ||
| Tensor::from_data( | ||
| self.data_weights.clone().convert::<PS::FloatElem>(), | ||
| &device, | ||
| ), | ||
| ), | ||
| bias: self.data_bias.as_ref().map(|bias| { | ||
| Param::initialized( | ||
| ParamId::new(), | ||
| Tensor::from_data(bias.clone().convert::<PS::FloatElem>(), &device), | ||
| ) | ||
| }), | ||
| stride: ConstantRecord::new(), | ||
| kernel_size: ConstantRecord::new(), | ||
| dilation: ConstantRecord::new(), | ||
| groups: ConstantRecord::new(), | ||
| padding: ConstantRecord::new(), | ||
| padding_out: ConstantRecord::new(), | ||
| channels: [ConstantRecord::new(); 2], | ||
| }; | ||
|
|
||
| let item = Record::into_item::<PS>(record); | ||
| item.serialize(serializer) | ||
| } | ||
|
|
||
| fn forward(&self, scope: &mut Scope, node_position: usize) -> TokenStream { | ||
| let input = scope.tensor_use_owned(&self.input, node_position); | ||
| let output = &self.output.name; | ||
| let field = &self.field.name; | ||
|
|
||
| quote! { | ||
| let #output = self.#field.forward(#input); | ||
| } | ||
| } | ||
| fn register_imports(&self, imports: &mut BurnImports) { | ||
| imports.register("burn::nn::conv::ConvTranspose1d"); | ||
| imports.register("burn::nn::conv::ConvTranspose1dConfig"); | ||
| } | ||
|
|
||
| fn into_node(self) -> Node<PS> { | ||
| Node::ConvTranspose1d(self) | ||
| } | ||
| } | ||
|
|
||
| #[cfg(test)] | ||
| mod tests { | ||
| use super::*; | ||
| use crate::burn::{ | ||
| graph::BurnGraph, | ||
| node::{conv_transpose_1d::ConvTranspose1dNode, test::assert_tokens}, | ||
| TensorType, | ||
| }; | ||
| use burn::{nn::conv::ConvTranspose1dConfig, record::FullPrecisionSettings}; | ||
|
|
||
| #[test] | ||
| fn test_codegen() { | ||
| let mut graph = BurnGraph::<FullPrecisionSettings>::default(); | ||
|
|
||
| graph.register(ConvTranspose1dNode::new( | ||
| "conv_transpose_1d", | ||
| TensorType::new_float("input", 3), | ||
| TensorType::new_float("output", 3), | ||
| TensorData::from([2f32]), | ||
| None, | ||
| ConvTranspose1dConfig::new([3, 3], 3).with_padding(0), | ||
| )); | ||
|
|
||
| graph.register_input_output(vec!["input".to_string()], vec!["output".to_string()]); | ||
|
|
||
| let expected = quote! { | ||
| use burn::{ | ||
| module::Module, | ||
| tensor::{backend::Backend, Tensor}, | ||
| }; | ||
| use burn::nn::conv::ConvTranspose1d; | ||
| use burn::nn::conv::ConvTranspose1dConfig; | ||
|
|
||
| #[derive(Module, Debug)] | ||
| pub struct Model <B: Backend> { | ||
| conv_transpose_1d: ConvTranspose1d<B>, | ||
| phantom: core::marker::PhantomData<B>, | ||
| device: burn::module::Ignored<B::Device>, | ||
| } | ||
|
|
||
| impl<B: Backend> Model <B> { | ||
| #[allow(unused_variables)] | ||
| pub fn new(device: &B::Device) -> Self { | ||
| let conv_transpose_1d = ConvTranspose1dConfig::new([3, 3], 3) | ||
| .with_stride(1) | ||
| .with_padding(0) | ||
| .with_padding_out(0) | ||
| .with_dilation(1) | ||
| .with_groups(1) | ||
| .with_bias(true) | ||
| .init(device); | ||
|
|
||
| Self { | ||
| conv_transpose_1d, | ||
| phantom: core::marker::PhantomData, | ||
| device: burn::module::Ignored(device.clone()), | ||
| } | ||
| } | ||
| #[allow(clippy::let_and_return, clippy::approx_constant)] | ||
| pub fn forward(&self, input: Tensor<B, 3>) -> Tensor<B, 3> { | ||
| let output = self.conv_transpose_1d.forward(input); | ||
|
|
||
| output | ||
| } | ||
| } | ||
| }; | ||
|
|
||
| assert_tokens(graph.codegen(), expected); | ||
| } | ||
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.