Please refer to [Installation](install_en.md) to setup environment at first, and prepare flower102 dataset by following the instruction mentioned in the [Quick Start](quick_start_en.md).
If training and evaluation are performed on CPU or single GPU, it is recommended to use the `tools/train.py` and `tools/eval.py`.
For training and evaluation in multi-GPU environment on Linux, please refer to [2. Training and evaluation on Linux+GPU](#2-training-and-evaluation-on-linuxgpu).
Among them, `-c` is used to specify the path of the configuration file, `-o` is used to specify the parameters needed to be modified or added, `-o pretrained_model=""` means to not using pre-trained models.
`-o use_gpu=True` means to use GPU for training. If you want to use the CPU for training, you need to set `use_gpu` to `False`.
Of course, you can also directly modify the configuration file to update the configuration. For specific configuration parameters, please refer to [Configuration Document](config_en.md).
* If mixup or cutmix is not used during training, in addition to the above information, top-1 and top-k (The default is 5) will also be printed in the log:
Among them, `-o pretrained_model` is used to set the address to load the pretrained weights. When using it, you need to replace it with your own pretrained weights' path, or you can modify the path directly in the configuration file.
We also provide a lot of pre-trained models trained on the ImageNet-1k dataset. For the model list and download address, please refer to the [model library overview](../models/models_intro_en.md).
The configuration file does not need to be modified. You only need to add the `checkpoints` parameter during training, which represents the path of the checkpoints. The parameter weights, learning rate, optimizer and other information will be loaded using this parameter.
**Note**:
* The parameter `-o last_epoch=5` means to record the number of the last training epoch as `5`, that is, the number of this training epoch starts from `6`, , and the parameter defaults to `-1`, which means the number of this training epoch starts from `0`.
* The `-o checkpoints` parameter does not need to include the suffix of the checkpoints. The above training command will generate the checkpoints as shown below during the training process. If you want to continue training from the epoch `5`, Just set the `checkpoints` to `./output/MobileNetV3_large_x1_0_gpupaddle/5/ppcls`, PaddleClas will automatically fill in the `pdopt` and `pdparams` suffixes.
The above command will use `./configs/quick_start/MobileNetV3_large_x1_0_finetune.yaml` as the configuration file to evaluate the model `./output/MobileNetV3_large_x1_0/best_model/ppcls`. You can also set the evaluation by changing the parameters in the configuration file, or you can update the configuration with the `-o` parameter, as shown above.
**Note:** If the model is a dygraph type, you only need to specify the prefix of the model file when loading the model, instead of specifying the suffix, such as [1.3 Resume Training](#13-resume-training).
If you want to run PaddleClas on Linux with GPU, it is highly recommended to use `paddle.distributed.launch` to start the model training script(`tools/train.py`) and evaluation script(`tools/eval.py`), which can start on multi-GPU environment more conveniently.
After preparing the configuration file, The training process can be started in the following way. `paddle.distributed.launch` specifies the GPU running card number by setting `selected_gpus`:
Among them, `pretrained_model` is used to set the address to load the pretrained weights. When using it, you need to replace it with your own pretrained weights' path, or you can modify the path directly in the configuration file.
There contains a lot of examples of model finetuning in [Quick Start](./quick_start_en.md). You can refer to this tutorial to finetune the model on a specific dataset.
The configuration file does not need to be modified. You only need to add the `checkpoints` parameter during training, which represents the path of the checkpoints. The parameter weights, learning rate, optimizer and other information will be loaded using this parameter. About `last_epoch` parameter, please refer [1.3 Resume training](#13-resume-training) for details.
+ `pre_label_image`: Whether to pre-label the image data, default value: `False`;
+ `pre_label_out_idr`: The output path of pre-labeled image data. When `pre_label_image=True`, a lot of subfolders will be generated under the path, each subfolder represent a category, which stores all the images predicted by the model to belong to the category.
**Note**: If you want to use `Transformer series models`, such as `DeiT_***_384`, `ViT_***_384`, etc., please pay attention to the input size of model, and need to set `resize_short=384`, `resize=384`.
Among them, the `--model` parameter is used to specify the model name, `--pretrained_model` parameter is used to specify the model file path, the path does not need to include the model file suffix name, and `--output_path` is used to specify the storage path of the converted model, class_dim means number of class for the model, default as 1000.
1. If `--output_path=./inference`, then three files will be generated in the folder `inference`, they are `inference.pdiparams`, `inference.pdmodel` and `inference.pdiparams.info`.
2. You can specify the `shape` of the model input image by setting the parameter `--img_size`, the default is `224`, which means the shape of input image is `224*224`.
The above command will generate the model structure file (`inference.pdmodel`) and the model weight file (`inference.pdiparams`), and then the inference engine can be used for inference:
+ `use_gpu`: Whether to use the GPU, default by `True`
+ `enable_mkldnn`: Wheter to use `MKL-DNN`, default by `False`. When both `use_gpu` and `enable_mkldnn` are set to `True`, GPU is used to run and `enable_mkldnn` will be ignored.
**Note**: If you want to use `Transformer series models`, such as `DeiT_***_384`, `ViT_***_384`, etc., please pay attention to the input size of model, and need to set `resize_short=384`, `resize=384`.
If you want to evaluate the speed of the model, it is recommended to use [predict.py](../../../tools/infer/predict.py), and enable TensorRT to accelerate.
