| MindData | MindSpore data framework, which provides data loading, enhancement, dataset management, and visualization. |
| MindInsight | MindSpore visualization component, which visualizes information such as scalars, images, computational graphs, and model hyperparameters. |
| MindSpore | Huawei-leaded open-source deep learning framework. |
| MindSpore Predict | A lightweight deep neural network inference engine that provides the inference function for models trained by MindSpore on the device side. |
| MindSpore Lite | A lightweight deep neural network inference engine that provides the inference function for models trained by MindSpore on the device side. |
| MNIST database | Modified National Handwriting of Images and Technology database, a large handwritten digit database, which is usually used to train various image processing systems. |
| ONNX | Open Neural Network Exchange, is an open format built to represent machine learning models.|
| PyNative Mode | MindSpore dynamic graph mode. In this mode, operators in the neural network are delivered and executed one by one, facilitating the compilation and debugging of the neural network model. |
Take the LeNet network as an example. The generated on-device model file is `lenet.pb`. The complete sample code `lenet.py` is as follows:
Take the LeNet network as an example. The generated on-device model file is `lenet.mindir`. The complete sample code `lenet.py` is as follows:
```python
importos
importnumpyasnp
...
...
@@ -171,9 +171,9 @@ To perform on-device model inference using MindSpore, perform the following step
else:
print("checkpoint file does not exist.")
```
3. In `mindspore/output/MSLite-0.6.0-linux_x86_64/converter` directory, calling MindSpore convert tool named `converter_lite`, convert model file (`.pb`) to on_device inference model file (`.ms`).
3. In `mindspore/output/MSLite-0.6.0-linux_x86_64/converter` directory, calling MindSpore convert tool named `converter_lite`, convert model file (`.mindir`) to on_device inference model file (`.ms`).
@@ -114,7 +114,7 @@ MindSpore supports the following inference scenarios based on the hardware platf
The Ascend 310 AI processor is equipped with the ACL framework and supports the OM format which needs to be converted from the model in ONNX or AIR format. For inference on the Ascend 310 AI processor, perform the following steps:
1. Generate a model in ONNX or AIR format on the training platform. For details, see [Export AIR Model and ONNX Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#aironnx).
1. Generate a model in ONNX or AIR format on the training platform. For details, see [Export AIR Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#export-air-model) and [Export ONNX Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#export-onnx-model).
2. Convert the ONNX or AIR model file into an OM model file and perform inference.
- For performing inference in the cloud environment (ModelArt), see the [Ascend 910 training and Ascend 310 inference samples](https://support.huaweicloud.com/bestpractice-modelarts/modelarts_10_0026.html).
...
...
@@ -128,7 +128,7 @@ The inference is the same as that on the Ascend 910 AI processor.
### Inference Using an ONNX File
1. Generate a model in ONNX format on the training platform. For details, see [Export AIR Model and ONNX Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#aironnx).
1. Generate a model in ONNX format on the training platform. For details, see [Export ONNX Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#export-onnx-model).
2. Perform inference on a GPU by referring to the runtime or SDK document. For example, use TensorRT to perform inference on the NVIDIA GPU. For details, see [TensorRT backend for ONNX](https://github.com/onnx/onnx-tensorrt).
...
...
@@ -140,10 +140,10 @@ The inference is the same as that on the Ascend 910 AI processor.
### Inference Using an ONNX File
Similar to the inference on a GPU, the following steps are required:
1. Generate a model in ONNX format on the training platform. For details, see [Export AIR Model and ONNX Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#aironnx).
1. Generate a model in ONNX format on the training platform. For details, see [Export ONNX Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#export-onnx-model).
2. Perform inference on a CPU by referring to the runtime or SDK document. For details about how to use the ONNX Runtime, see the [ONNX Runtime document](https://github.com/microsoft/onnxruntime).
## On-Device Inference
MindSpore Predict is an inference engine for on-device inference. For details, see[On-Device Inference](https://www.mindspore.cn/tutorial/en/master/advanced_use/on_device_inference.html).
MindSpore Lite is an inference engine for on-device inference. For details, see [Export MINDIR Model](https://www.mindspore.cn/tutorial/en/master/use/saving_and_loading_model_parameters.html#export-mindir-model) and[On-Device Inference](https://www.mindspore.cn/tutorial/en/master/advanced_use/on_device_inference.html).
-[Export AIR Model and ONNX Model](#export-air-model-and-onnx-model)
-[Export Model](#export-model)
-[Export AIR Model](#export-air-model)
-[Export ONNX Model](#export-onnx-model)
-[Export MINDIR Model](#export-mindir-model)
<!-- /TOC -->
...
...
@@ -140,10 +143,14 @@ model.train(epoch, dataset)
The `load_checkpoint` method returns a parameter dictionary and then the `load_param_into_net` method loads parameters in the parameter dictionary to the network or optimizer.
## Export AIR Model and ONNX Model
When you have a CheckPoint file, if you want to do inference, you need to generate corresponding models based on the network and CheckPoint.
Currently we support the export of AIR models based on Ascend AI processor and the export of ONNX models. Taking the export of AIR model as an example to illustrate the implementation of model export,
the code is as follows:
## Export Model
When you have a CheckPoint file, if you want to do inference, you need to generate corresponding models based on the network and CheckPoint. The `export` interface supports exporting multiple types of model file formats for inference on different hardware platforms.
### Export AIR Model
AIR format file only supports Ascend AI processor. The code example of exporting this format file is as follows:
Before using the `export` interface, you need to import` mindspore.train.serialization`.
The `input` parameter is used to specify the input shape and data type of the exported model.
If you want to export the ONNX model, you only need to specify the `file_format` parameter in the` export` interface as ONNX: `file_format = 'ONNX'`.
## Export MINDIR Model
It is recommended to use '.air' as the suffix of AIR format files.
### Export ONNX Model
ONNX format file is a general model file, which can be applied to many kinds of hardware, such as Ascend AI processor, GPU, CPU, etc. The code example of exporting this format file is as follows:
It is recommended to use '.onnx' as the suffix of ONNX format files.
### Export MINDIR Model
MINDIR format file can be applied to MindSpore Lite and MindSpore Serving. Currently, it supports inference network based on static graph without control flow semantics.
If you want to do inference on the device, then you need to generate corresponding MINDIR models based on the network and CheckPoint.
Currently we support the export of MINDIR models for inference based on graph mode, which don't contain control flow. Taking the export of MINDIR model as an example to illustrate the implementation of model export,
