Migrating From Third Party Frameworks With MindConverter

Overview

MindConverter is a migration tool to transform the model file of PyTorch(ONNX) or TensorFlow(PB) to MindSpore. The model file contains model structure definition(network) and weights information(weights), which will be transformed into model scripts(model.py) and weights file(ckpt) in MindSpore.

mindconverter-overview

Moreover, this tool is able to transform the model file of PyTorch to MindSpore by adding API(pytorch2mindspore) to original PyTorch scripts.

Quick Starts

MindConverter installation is described in Install Tools. After the installation, you will have the MindConverter CLI and MindConverter API with the basic usage as follows.

Start with MindConverter CLI

mindconverter --model_file /path/to/model_file --shape SHAPE --input_nodes INPUTS --output_nodes OUTPUTS

--model_file specifies the path of model file, the model file supports onnx format or pb format.
--shape specifies the input shape of model. Multiple inputs are separated by space.
--input_nodes specifies the input names of model. Multiple inputs are separated by space.
--output_nodes specifies the output names of model. Multiple outputs are separated by space.
Output files are generated and saved under $PWD/output by default.

Notes:

The model file is in onnx format. If the model input shape is a static value, just specify -- model_ file to complete the conversion. Otherwise, you need to specify --shape and --input_nodes to complete the conversion. --output_nodes can be omitted. For model input shape judgment, please refer to Determine the form of model input shape.
The model file is in pb format without special scenarios.

For more CLI arguments, please refer to MindConverter CLI Usage.

Start with MindConverter API

Write the following code in PyTorch network scripts.

from mindconverter import pytorch2mindspore
pytorch2mindspore(model, dummy_inputs)

For the usage of API, please refer to MindConverter API Description.

Install Tools

Install Dependencies

The following dependencies are required for MindConverter usage and suggested to be installed under the x86 environment. Refer to installation under the ARM environment.

Start with MindConverter CLI

# Corresponding version of MindSpore is required(e.g. r1.2).
pip install mindspore~=1.2.0

# ONNX and relevant libraries are required.
pip install onnx~=1.8.0
pip install onnxoptimizer~=0.1.2
pip install onnxruntime~=1.5.2

# tf2onnx is required if model file is in pb format.
pip install tf2onnx~=1.7.1

Start with MindConverter API

# Corresponding version of MindSpore is required(e.g. r1.6).
pip install mindspore~=1.6.0

# torch is required, and official LTS version 1.8.2 is recommended.
pip install torch~=1.8.2+cpu -f https://download.pytorch.org/whl/lts/1.8/torch_lts.html

Install MindConverter

You can install MindConverter either by pip or by source code.

Installation by pip

Install from PyPI:

pip install mindconverter

Install with customized version:

pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/{version}/MindInsight/any/mindconverter-{version}-py3-none-any.whl --trusted-host ms-release.obs.cn-north-4.myhuaweicloud.com -i https://pypi.tuna.tsinghua.edu.cn/simple

When the network is connected, dependency items are automatically downloaded during .whl package installation. (For details about other dependency items, see requirements.txt). In other cases, you need to manually install dependency items.

{version} denotes the version of MindConverter. For example, when you are downloading MindConverter 1.6.0, {version} should be 1.6.0.

MindConverter supports only Linux distro with x86 architecture 64-bit or ARM architecture 64-bit.

Installation by Source Code

Downloading Source Code from Gitee.

git clone https://gitee.com/mindspore/mindinsight.git -b r1.6

Compiling MindConverter by any of the following installation methods:

Run the following command in the root directory of the source code:

cd mindinsight/ecosystem_tools/mindconverter
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
python setup.py install

Build the whl package for installation.

Enter the root directory of the source code, first execute the MindConverter compilation script in the build directory, and then execute the command to install the whl package generated in the output directory.
```
cd mindinsight/ecosystem_tools/mindconverter
bash build/build.sh mindconverter
pip install output/mindconverter-{version}-py3-none-any.whl -i https://pypi.tuna.tsinghua.edu.cn/simple
```

Migration Solution

A typical model project contains 4 main components. Tips for migrating each component are as follows:

Model definition（model.py）
1. Transform the model structure with MindConverter CLI.
2. Manually enhance the readability of the generated model scripts(Optional).
3. Mix the generated model with the original project to validate equivalence of the migration. Refer to FAQ.
Data processing（dataset.py）
1. For a built-in dataset, please query API mapping for migration.
2. For a customized dataset and data augmentation, please refer to the migration template.
Model training（train.py）
1. The loss function(loss_fn) can be migrated by querying API mapping or user’s implementation.
2. The optimizer(optimizer) can be migrated by querying API mapping or user’s implementation.
3. As the training codes could be flexible and significantly different from MindSpore, implementation by the users is recommended. Please refer to FAQ.
Model evaluation（eval.py）
1. The metric(metric) can be migrated by querying API mapping or user’s implementation.
2. As the evaluation codes could be flexible and significantly different from MindSpore, implementation by the users is recommended. Please refer to FAQ.

Practice Guidance

Step 0：Export the model file

Exporting ONNX model file from PyTorch model(refer to FAQ for Tensorflow guidance) requires operators mapping between PyTorch and ONNX. Guidance is as follows:

Download source codes, weights file and relevant dataset files of the model project.
Dive into the model definition. Make sure that all parameters of the forward function are Tensor type. Please refer to FAQ.

Locate the model object and the input shape information from the model evaluation. Export the model object into onnx format.

import torch

# Replace the following classpath according to the actual situation.
from customized.path.to.pytorch.model import PyTorchNetwork

model = PyTorchNetwork()
param_dict = torch.load('/path/to/weights.pth')
model.load_state_dict(param_dict)

input_shape = (1, 3, 224, 224)
input_tensor = torch.randn(*input_shape, dtype=torch.float32)
torch.onnx.export(model, input_tensor, '/path/to/model.onnx', opset_version=11)

Validate the equivalence onnx model file against the original scripts.

import onnxruntime
import numpy as np

session = onnxruntime.InferenceSession('/path/to/model.onnx')
input_node = session.get_inputs()[0]
output = session.run(None, {input_node.name: np.load('/path/to/input.npy')})
assert np.allclose(output, np.load('/path/to/output.npy'))

Step 1:Migrate the model definition

Run the following MindConverter CLI to generate the model scripts(model.py), weights information(ckpt), migration reports and weights mapping.

mindconverter --model_file /path/to/model.onnx

If you need to specify the model input shape, input node names and output node names. Netron is recommended to get the above information.

Model scripts(model.py) and weights information(ckpt) can be used not only to validate the equivalence of migration, but also to generate the MindIR file.

import mindspore
import numpy as np

# Replace the following classpath according to the actual situation.
from customized.path.to.mindspore.model import MindSporeNetwork

network = MindSporeNetwork()
param_dict = mindspore.load_checkpoint('network.ckpt')
mindspore.load_param_into_net(network, param_dict)

input_data = np.load('/path/to/input.npy')
output_benchmark = np.load('/path/to/output.npy')

# Validate the equivalence of migration.
output_data = network(mindspore.Tensor(input_data))
assert np.allclose(output_data.asnumpy(), output_benchmark)

# Generate the MindIR file.
mindspore.export(network, mindspore.Tensor(input_data)), file_name='your_network_name', file_format='MINDIR')

Notes:

The Dropout operator will be lost after conversion because the inference mode is used to load the ONNX or TensorFlow model. Manually re-implementation is necessary.
This script conversion tool relies on operators which supported by MindConverter and MindSpore. Unsupported operators may not be successfully mapped to MindSpore operators. You can manually edit, or implement the mapping based on MindConverter, and make contribution to our MindInsight repository. We appreciate your support for the MindSpore community.
MindConverter converts dynamic input shape to constant one based on --shape while using graph based scheme, as a result, it is required that inputs’ shape used to retrain or inference in MindSpore are the same as that used to convert using MindConverter. If the input shape has changed, please running MindConverter again with new --shape or fixing shape related parameters in the old script.
MindSpore script and MindSpore checkpoint file are saved in the one file folder path, while report file and weight map file are saved in the other one.
The security and consistency of the model file should be guaranteed by the user.

Step 2：Migrate the data processing

For a built-in dataset, please query API mapping for migration. For a customized dataset and data augmentation, self implementation is recommended. For more data processing migration, please refer to the programming guidance.

Source codes with PyTorch framework are as follows:

from torch.utils.data import Dataset, DataLoader
from torchvision import transforms

class CustomDataset(Dataset):
    def __init__(self, *args, **kwargs):
        self.records = [...]
        self.labels = [...]
        # Define data augmentation.
        self.transform = transforms.Compose([
            transforms.ToTensor(),
            transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
        ])
    def __len__(self):
        return len(self.labels)
    def __getitem__(self, idx):
        # Execute data augmentation.
        data = self.transform(self.records[idx])
        return data, self.labels[idx]

dataset = CustomDataset(*args, **kwargs)
data_loader = DataLoader(dataset, batch_size=BATCH_SIZE)

Corresponding generated codes with MindSpore framework are as follows:

from mindspore.dataset import GeneratorDataset
from mindspore.dataset import py_transforms as transforms

class CustomGenerator:
    def __init__(self, *args, **kwargs):
        self.records = [...]
        self.labels = [...]
        # Define data augmentation.
        self.transform = transforms.Compose([
            transforms.ToTensor(),
            transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
        ])
    def __len__(self):
        return len(self.labels)
    def __getitem__(self, idx):
        # Execute data augmentation.
        data = self.transform(self.records[idx])
        return data, self.labels[idx]

generator = CustomGenerator(*args, **kwargs)
dataset = GeneratorDataset(generator, column_names=['data', 'label']).batch(BATCH_SIZE)

Step 3：Migrate the model training

The loss function(loss_fn) can be migrated by querying API mapping or user’s implementation. For more loss function migration, please refer to the programming guidance.

The optimizer(optimizer) can be migrated by querying API mapping or user’s implementation. For more optimizer migration, please refer to the programming guidance.

As the training codes could be flexible and significantly different from MindSpore, implementation by the users is recommended.

Source codes with PyTorch framework are as follows:

import torch

# Replace the following classpath according to the actual situation.
from customized.path.to.pytorch.model import PyTorchNetwork

# Create a instance of network model.
network = PyTorchNetwork()

# Define optimizer and learning rate.
optimizer = torch.optim.SGD(network.parameters(), lr=LEARNING_RATE)
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=DECAY_RATE)

# Launch the model training.
for i in range(EPOCH_SIZE):
    for data, label in data_loader:
        optimizer.zero_grad()
        output = network(data)
        loss = loss_fn(output, label)
        loss.backward()
        optimizer.step()
    scheduler.step()

Corresponding generated codes(Low-Level API) with MindSpore framework are as follows:

from mindspore import nn

# Replace the following classpath according to the actual situation.
from customized.path.to.mindspore.model import MindSporeNetwork

# Create a instance of network model.
network = MindSporeNetwork()

# Define learning rate and optimizer.
scheduler = nn.ExponentialDecayLR(LEARNING_RATE, decay_rate=DECAY_RATE, decay_steps=1)
optimizer = nn.SGD(params=network.trainable_params(), learning_rate=scheduler)

# Launch the model training.
net_with_loss = nn.WithLossCell(network, loss_fn)
train_network = nn.TrainOneStepCell(net_with_loss, optimizer)
train_network.set_train()
data_iterator = dataset.create_tuple_iterator(num_epochs=EPOCH_SIZE)
for i in range(EPOCH_SIZE):
    for data, label in data_iterator:
        loss = train_network(data, label)

Corresponding generated codes(High-Level API) with MindSpore framework are as follows:

from mindspore import nn
from mindspore import Model

# Replace the following classpath according to the actual situation.
from customized.path.to.mindspore.model import MindSporeNetwork

# Create a instance of network model.
network = MindSporeNetwork()

# Define learning rate and optimizer.
scheduler = nn.ExponentialDecayLR(LEARNING_RATE, decay_rate=DECAY_RATE, decay_steps=1)
optimizer = nn.SGD(params=network.trainable_params(), learning_rate=scheduler)

# Launch the model training.
model = Model(network, loss_fn=loss_fn, optimizer=optimizer)
model.train(EPOCH_SIZE, dataset)

Step 4：Migrate the model evaluation

The metric(metric) can be migrated by querying API mapping or user’s implementation.

As the evaluation codes could be flexible and significantly different from MindSpore, implementation by the users is recommended. For more model evaluation migration, please refer to the programming guidance.

Source codes with PyTorch framework are as follows:

import torch

# Replace the following classpath according to the actual situation.
from customized.path.to.pytorch.model import PyTorchNetwork

network = PyTorchNetwork()
param_dict = torch.load('/path/to/weights.path')
network.load_state_dict(param_dict)


for data, label in data_loader:
    output = network(data)
    loss = loss_fn(output, label)
    accuracy = metric_fn(output, label)
    print(accuracy)

Corresponding generated codes(Low-Level API) with MindSpore framework are as follows:

import mindspore

# Replace the following classpath according to the actual situation.
from customized.path.to.mindspore.model import MindSporeNetwork

network = MindSporeNetwork()
param_dict = mindspore.load_checkpoint('/path/to/weights.ckpt')
mindspore.load_param_into_net(network, param_dict)

data_iterator = dataset.create_tuple_iterator()
for data, label in data_iterator:
    output = network(data)
    loss = loss_fn(output, label)
    accuracy = metric_fn(output, label)
    print(accuracy)

Corresponding generated codes(High-Level API) with MindSpore framework are as follows:

import mindspore
from mindspore import Model

# Replace the following classpath according to the actual situation.
from customized.path.to.mindspore.model import MindSporeNetwork

network = MindSporeNetwork()
param_dict = mindspore.load_checkpoint('/path/to/weights.ckpt')
mindspore.load_param_into_net(network, param_dict)

model = Model(network, loss_fn=loss_fn, metrics={'accuracy'})
accuracy = model.eval(dataset)

MindConverter CLI Usage

usage: mindconverter [-h] [--version]
                     [--model_file MODEL_FILE] [--shape SHAPE [SHAPE ...]]
                     [--input_nodes INPUT_NODES [INPUT_NODES ...]]
                     [--output_nodes OUTPUT_NODES [OUTPUT_NODES ...]]
                     [--output OUTPUT] [--report REPORT]

Arguments are as follows：

Argument	Mandatory	Description	Type	Default	Example
-h, –help	N	Show the help message.	-	-	-
–version	N	Show the version info.	-	-	-
–model_file	Y	Specify the path of model file.	String	-	/path/to/model.onnx
–shape	Y	Specify the input shape of model. Multiple inputs are separated by space.	String	-	1,3,224,224
–input_nodes	Y	Specify the input names of model. Multiple inputs are separated by space.	String	-	input_1:0
–output_nodes	Y	Specify the output names of model. Multiple outputs are separated by space.	String	-	output_1:0 output_2:0
–output	N	Specify the directory path for generated files.	String	$PWD	/path/to/output/dir

Model List Supported by MindConverter

For supported models (tested based on PyTorch 1.5.0 and TensorFlow 1.15.0, x86 Ubuntu released version), please refer to LINK.

MindConverter Error Code Definition

For error code defined in MindConverter, please refer to LINK.

FAQ

Install dependencies under ARM

MindConverter under the ARM environment requires compiling protobuf/onnx/onnxoptimizer from scratch. Since the compiling is complicated and error prone, we strongly recommend to use MindConverter under the x86 environment.

Compile protobuf(refer to ONNX) and install cpp backend.

# build, compile and install protobuf
git clone https://github.com/protocolbuffers/protobuf.git
cd protobuf
git checkout v3.16.0
git submodule update --init --recursive
mkdir build_source
cd build_source
cmake ../cmake -Dprotobuf_BUILD_SHARED_LIBS=OFF -DCMAKE_INSTALL_PREFIX=/usr/local/protobuf -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_POSITION_INDEPENDENT_CODE=ON -Dprotobuf_BUILD_TESTS=OFF -DCMAKE_BUILD_TYPE=Release
make -j$(nproc)
make install

# Install cpp backend.
cd ../python
python setup.py install --cpp_implementation

Configure environment variables for protobuf.

export PROTOBUF_PATH=/usr/local/protobuf
export PATH=$PROTOBUF_PATH/bin:$PATH
export PKG_CONFIG_PATH=$PROTOBUF_PATH/lib/pkgconfig
export LD_LIBRARY_PATH=$PROTOBUF_PATH/lib:$LD_LIBRARY_PATH
export LIBRARY_PATH=$PROTOBUF_PATH/lib:$LIBRARY_PATH
export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=cpp

Validate cpp backend of protobuf.

from google.protobuf.internal import api_implementation
print(api_implementation.Type())

onnx should be recompiled and installed to guarantee running onnx with protobuf built by static library. Please refer to the installation guidance.

git clone https://github.com/onnx/onnx.git
cd onnx
git submodule update --init --recursive
# prefer lite proto
set CMAKE_ARGS=-DONNX_USE_LITE_PROTO=ON
pip install -e .

Compile and install onnxoptimizer. Please refer to the installation guidance.

git clone --recursive https://github.com/onnx/optimizer onnxoptimizer
cd onnxoptimizer
pip3 install -e .

Install onnxruntime.
```
pip install onnxruntime~=1.5.2
```

Determine the form of model input shape

Using Netron load the onnx model file, click the top node, and observe the type of each input in INPUTS in the sidebar. If the type is a specific value, such as int64[1,9], the current input is static. Otherwise, it is dynamic, such as int64[batch,sequence].

Export the model file of Tensorflow

Exporting the PB model file from a Tensorflow model requires operators mapping between Tensorflow and ONNX. For models defined by Keras, guidance is as follows:

TensorFlow 1.x

import tensorflow as tf
from tensorflow.python.framework import graph_io
from tensorflow.python.keras.applications.inception_v3 import InceptionV3

model = InceptionV3()
INPUT_NODES = [ipt.op.name for ipt in model.inputs]
OUTPUT_NODES = [opt.op.name for opt in model.outputs]

tf.keras.backend.set_learning_phase(0)
session = tf.keras.backend.get_session()
with session.graph.as_default():
    graph_inf = tf.graph_util.remove_training_nodes(session.graph.as_graph_def())
    graph_frozen = tf.graph_util.convert_variables_to_constants(session, graph_inf, OUTPUT_NODES)
    graph_io.write_graph(graph_frozen, logdir="/path/to/output/dir", name="model.pb", as_text=False)

print(f"Input nodes name: {INPUT_NODES}, output nodes name: {OUTPUT_NODES}")

TensorFlow 2.x

import tensorflow as tf
from tensorflow.python.framework.convert_to_constants import convert_variables_to_constants_v2
from tensorflow.keras.applications import InceptionV3

model = InceptionV3()
spec = tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype)
full_model = tf.function(lambda x: model(x)).get_concrete_function(spec)
frozen_func = convert_variables_to_constants_v2(full_model)
frozen_func.graph.as_graph_def()
tf.io.write_graph(frozen_func.graph, logdir="/path/to/output/dir", name="model.pb", as_text=False)

TensorFlow is required for exporting PB model file but it is not explicitly declared as mandatory dependency for MindConverter. If the user wants to use graph based MindConverter, please install TensorFlow(TensorFlow 1.15.x is recommended).

Rectify parameters of forward function definition

Some models define non-Tensor parameters within forward function are as follows:

class Model(nn.Cell):
    def __init__(self, *args, **kwargs):
        self.op = Operator()
        self.loss_fn = LossFunction()
    def forward(self, data, label):
        output = self.op(data)
        loss = self.loss_fn(output, label)
        return output, loss

The above label is a non-Tensor parameter which needs to be rectified.

class Model(nn.Cell):
    def __init__(self, *args, **kwargs):
        self.op = Operator()
    def forward(self, data):
        output = self.op(data)
        return output

Mix the MindSpore model with the original training scripts

Validate the equivalence of migration by mixing the MindSpore model and weights with the PyTorch training scripts.

import mindspore
import torch
from torch.utils.data import DataLoader

# Replace the following classpath according to the actual situation.
from customized.path.to.mindspore.model import MindSporeNetwork

network = MindSporeNetwork()
param_dict = mindspore.load_checkpoint('network.ckpt')
mindspore.load_param_into_net(network, param_dict)

data_loader = DataLoader(dataset, batch_size=BATCH_SIZE)
for data, label in data_loader:
    ms_data = mindspore.Tensor(data.numpy())
    ms_output = network(ms_data)
    output = torch.Tensor(ms_output.asnumpy())
    loss = loss_fn(output, label)
    accuracy = metric_fn(output, label)
    print(accuracy)

Migration reports and weights mapping

For operators that are not successfully converted, the conversion report records the unconverted code lines and operator information, and at the same time identifies the input/output shape of the node in the code (represented as input_shape and output_shape), which is convenient for users to modify manually. An example of the Reshape operator is as follows:

class Classifier(nn.Cell):
    def __init__(self):
        super(Classifier, self).__init__()
        self.reshape = onnx.Reshape(input_shape=(1, 1280, 1, 1), output_shape=(1, 1280))

    def construct(self, x):
        # Suppose input of `reshape` is x.
        reshape_output = self.reshape(x)
        # skip codes ...

It is convenient to replace the operators according to the input_shape and output_shape parameters. The replacement is like this:

from mindspore import ops

class Classifier(nn.Cell):
    def __init__(self):
        super(Classifier, self).__init__()
        self.reshape = ops.Reshape(input_shape=(1, 1280, 1, 1), output_shape=(1, 1280))

    def construct(self, x):
        # Suppose input of `reshape` is x.
        reshape_output = self.reshape(x, (1, 1280))
        # skip codes ...

Weight information in MindSpore(converted_weight) and that in source framework(source_weight) are saved in weight mapping. An example is as follows:

{
  "resnet50": [
    {
      "converted_weight": {
        "name": "conv2d_0.weight",
        "shape": [64, 3, 7, 7],
        "data_type": "Float32"
      },
      "source_weight": {
        "name": "conv1.weight",
        "shape": [64, 3, 7, 7],
        "data_type": "float32"
      }
    }
  ]
}

AST-Based Model Migration

MindConverter supports AST-based model migration for PyTorch scripts. It parses and analyzes original scripts, then replaces them with the MindSpore AST to generate codes.

Since the result may differ due to the coding style of original scripts, AST-based model migration is now DEPRECATED and will be removed in r2.0.

Assume the PyTorch script is located at /path/to/model.py, and outputs the transformed MindSpore script to /path/to/output/dir. Use the following command:

mindconverter --in_file /path/to/model.py --output /path/to/output/dir

In the conversion report, non-converted code is listed as follows. x, y indicates the line number and the column number of the original scripts. For non-converted operators, please refer to MindSpore API mapping. For unsupported operators, the corresponding code lines will remain in the original way.

line x:y: [UnConvert] 'operator' didn't convert. ...

For non-converted operators, suggestions are provided in the report. For example, MindConverter suggests that replace torch.nn.AdaptiveAvgPool2d with mindspore.ops.ReduceMean in line 157:23.

 [Start Convert]
 [Insert] 'from mindspore import ops' is inserted to the converted file.
 line 1:0: [Convert] 'import torch' is converted to 'import mindspore'.
 ...
 line 157:23: [UnConvert] 'nn.AdaptiveAvgPool2d' didn't convert. Maybe could convert to mindspore.ops.ReduceMean.
 ...
 [Convert Over]

The following cases are not supported:

Specific classes and functions.
- torch.Tensor members, including shape，ndim and dtype
- torch.nn.AdaptiveXXXPoolXd and torch.nn.functional.adaptive_XXX_poolXd()
- torch.nn.functional.Dropout
- torch.unsqueeze() and torch.Tensor.unsqueeze()
- torch.chunk() and torch.Tensor.chunk()

Subclasses of nn.Module.

# Code snippets from torchvision.models.mobilenet
from torch import nn

class ConvBNReLU(nn.Sequential):
   def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, groups=1):
       padding = (kernel_size - 1) // 2
       super(ConvBNReLU, self).__init__(
           nn.Conv2d(in_planes, out_planes, kernel_size, stride, padding, groups=groups, bias=False),
           nn.BatchNorm2d(out_planes),
           nn.ReLU6(inplace=True)
       )