[Feature] add model script, training recipe and training weights of DeiT

configs/deit/README.md

@@ -0,0 +1,86 @@
+
+# DeiT
+> [Training data-efficient image transformers & distillation through attention](https://arxiv.org/abs/2012.12877)
+
+## Introduction
+
+DeiT: Data-efficient Image Transformers
+
+## Results
+
+**Implementation and configs for training were taken and adjusted from [this repository](https://gitee.com/cvisionlab/models/tree/deit/release/research/cv/DeiT), which implements Twins models in mindspore.**
+
+Our reproduced model performance on ImageNet-1K is reported as follows.
+
+<div align="center">
+
+| Model    | Context  | Top-1 (%) | Top-5 (%) | Params (M) | Recipe                                                                                        | Download                                                                               |
+|----------|----------|-----------|-----------|------------|-----------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------|
+| deit_base | Converted from PyTorch | 81.62     | 95.58     | -  | [yaml](https://github.com/mindspore-lab/mindcv/blob/main/configs/deit/deit_b_gpu.yaml) | [weights](https://storage.googleapis.com/huawei-mindspore-hk/DeiT/Converted/deit_base_patch16_224.ckpt) |
+| deit_base | 8xRTX3090 | 72.29 | 89.93 | - | [yaml](https://github.com/mindspore-lab/mindcv/blob/main/configs/deit/deit_b_gpu.yaml) | [weights](https://storage.googleapis.com/huawei-mindspore-hk/DeiT/deit_base_patch16_224_acc%3D0.725.ckpt)
+| deit_small | Converted from PyTorch | 79.39 | 94.80 | - | [yaml](https://github.com/mindspore-lab/mindcv/blob/main/configs/deit/deit_b_gpu.yaml) | [weights](https://storage.googleapis.com/huawei-mindspore-hk/DeiT/Converted/deit_small_patch16_224.ckpt) |
+| deit_tiny | Converted from PyTorch | 71.58 | 90.76 | - | [yaml](https://github.com/mindspore-lab/mindcv/blob/main/configs/deit/deit_b_gpu.yaml) | [weights](https://storage.googleapis.com/huawei-mindspore-hk/DeiT/Converted/deit_tiny_patch16_224.ckpt) |
+
+</div>
+
+#### Notes
+
+- Context: The weights in the table were taken from [official repository](https://github.com/facebookresearch/deit) and converted to mindspore format
+- Top-1 and Top-5: Accuracy reported on the validation set of ImageNet-1K.
+
+## Quick Start
+
+### Preparation
+
+#### Installation
+Please refer to the [installation instruction](https://github.com/mindspore-ecosystem/mindcv#installation) in MindCV.
+
+#### Dataset Preparation
+Please download the [ImageNet-1K](https://www.image-net.org/challenges/LSVRC/2012/index.php) dataset for model training and validation.
+
+### Training
+
+* Distributed Training
+
+
+```shell
+# distrubted training on multiple GPU/Ascend devices
+mpirun -n 8 python train.py --config configs/deit/deit_b_gpu.yaml --data_dir /path/to/imagenet --distributed True
+```
+
+> If the script is executed by the root user, the `--allow-run-as-root` parameter must be added to `mpirun`.
+
+Similarly, you can train the model on multiple GPU devices with the above `mpirun` command.
+
+For detailed illustration of all hyper-parameters, please refer to [config.py](https://github.com/mindspore-lab/mindcv/blob/main/config.py).
+
+**Note:**  As the global batch size  (batch_size x num_devices) is an important hyper-parameter, it is recommended to keep the global batch size unchanged for reproduction or adjust the learning rate linearly to a new global batch size.
+
+* Standalone Training
+
+If you want to train or finetune the model on a smaller dataset without distributed training, please run:
+
+```shell
+# standalone training on a CPU/GPU/Ascend device
+python train.py --config configs/deit/deit_b_gpu.yaml --data_dir /path/to/dataset --distribute False
+```
+
+### Validation
+
+To validate the accuracy of the trained model, you can use `validate.py` and parse the checkpoint path with `--ckpt_path`.
+
+```shell
+python validate.py -c configs/deit/deit_b_gpu.yaml --data_dir /path/to/imagenet --ckpt_path /path/to/ckpt
+```
+
+### Deployment
+
+Please refer to the [deployment tutorial](https://github.com/mindspore-lab/mindcv/blob/main/tutorials/deployment.md) in MindCV.
+
+## References
+
+Paper - https://arxiv.org/pdf/2012.12877.pdf
+
+Official repo - https://github.com/facebookresearch/deit
+
+Mindspore implementation - https://gitee.com/cvisionlab/models/tree/deit/release/research/cv/DeiT

configs/deit/deit_b_gpu.yaml

@@ -0,0 +1,72 @@
+# system
+mode: 0
+distribute: False
+num_parallel_workers: 4
+val_while_train: True
+
+# dataset
+dataset: 'imagenet'
+data_dir: 'path/to/imagenet/'
+shuffle: True
+dataset_download: False
+batch_size: 64
+drop_remainder: True
+train_split: val/
+val_split: val/
+
+# augmentation
+image_resize: 224
+scale: [0.08, 1.0]
+ratio: [0.75, 1.333]
+hflip: 0.5
+auto_augment: 'randaug-m9-mstd0.5-inc1'
+interpolation: bicubic
+re_prob: 0.25
+re_value: 'random'
+cutmix: 1.0
+mixup: 0.8
+mixup_prob: 1.0
+mixup_mode: batch
+mixup_off_epoch: 0
+switch_prob: 0.5
+crop_pct: 0.875
+color_jitter: 0.3
+
+# model
+model: 'deit_base'
+num_classes: 1000
+pretrained: False
+ckpt_path: ''
+
+keep_checkpoint_max: 10
+ckpt_save_dir: './ckpt'
+
+epoch_size : 300
+dataset_sink_mode: True
+amp_level: O2
+ema: False
+ema_decay: 0.99996
+clip_grad: False
+clip_value: 5.0
+
+# loss
+loss: 'CE'
+label_smoothing: 0.1
+
+# lr scheduler
+lr_scheduler: 'cosine_decay'
+min_lr: 1.0e-5
+lr: 0.0005
+warmup_epochs: 5
+warmup_factor: 0.002
+decay_epochs: 30
+decay_rate: 0.1
+
+# optimizer
+opt: 'adamw'
+filter_bias_and_bn: True
+eps: 1.0e-8
+momentum: 0.9
+weight_decay: 0.05
+dynamic_loss_scale: True
+use_nesterov: False

mindcv/models/__init__.py

@@ -3,6 +3,7 @@
     bit,
     convit,
     convnext,
+    deit,
     densenet,
     dpn,
     edgenext,
@@ -45,6 +46,7 @@
 from .bit import *
 from .convit import *
 from .convnext import *
+from .deit import *
 from .densenet import *
 from .dpn import *
 from .edgenext import *
@@ -91,6 +93,7 @@
 __all__.extend(bit.__all__)
 __all__.extend(convit.__all__)
 __all__.extend(convnext.__all__)
+__all__.extend(deit.__all__)
 __all__.extend(densenet.__all__)
 __all__.extend(dpn.__all__)
 __all__.extend(edgenext.__all__)