### Install Dependencies

Source: https://github.com/huggingface/pytorch-image-models/blob/main/hfdocs/README.md

Install the necessary packages for building Hugging Face documentation and for code formatting.

```bash
pip install git+https://github.com/huggingface/doc-builder.git@main#egg=hf-doc-builder
pip install watchdog black
```

--------------------------------

### Install huggingface_hub

Source: https://github.com/huggingface/pytorch-image-models/blob/main/hfdocs/source/hf_hub.mdx

Install the necessary package for Hugging Face Hub integration.

```bash
pip install huggingface_hub
```

--------------------------------

### Install timm using pip

Source: https://github.com/huggingface/pytorch-image-models/blob/main/hfdocs/source/installation.mdx

Install the timm library using pip for the most straightforward installation.

```bash
pip install timm
```

--------------------------------

### EfficientNet-B2 Training with RandAugment

Source: https://github.com/huggingface/pytorch-image-models/blob/main/hfdocs/source/training_script.mdx

Example training command for EfficientNet-B2 with RandAugment, optimized for dual Titan RTX cards with NVIDIA Apex installed. This configuration achieved 80.4% top-1 accuracy.

```bash
./distributed_train.sh 2 --data-dir /imagenet/ --model efficientnet_b2 -b 128 --sched step --epochs 450 --decay-epochs 2.4 --decay-rate .97 --opt rmsproptf --opt-eps .001 -j 8 --warmup-lr 1e-6 --weight-decay 1e-5 --drop 0.3 --drop-path 0.2 --model-ema --model-ema-decay 0.9999 --aa rand-m9-mstd0.5 --remode pixel --reprob 0.2 --amp --lr .016
```

--------------------------------

### Install Project Dependencies

Source: https://github.com/huggingface/pytorch-image-models/blob/main/CONTRIBUTING.md

Installs the project's main dependencies and development/testing dependencies using pip.

```bash
python -m pip install -r requirements.txt
```

```bash
python -m pip install -r requirements-dev.txt  # for testing
```

```bash
python -m pip install -e .
```

--------------------------------

### Load Pretrained Wide ResNet Model

Source: https://github.com/huggingface/pytorch-image-models/blob/main/hfdocs/source/models/wide-resnet.mdx

Loads a pretrained Wide ResNet model. Use this to quickly get started with image classification tasks.

```python
>>> import timm
>>> model = timm.create_model('wide_resnet101_2', pretrained=True)
>>> model.eval()
```

--------------------------------

### Create and navigate to project directory

Source: https://github.com/huggingface/pytorch-image-models/blob/main/hfdocs/source/installation.mdx

Use these commands to create a new project directory and navigate into it.

```bash
mkdir ~/my-project
cd ~/my-project
```

--------------------------------

### Create Model with No Classifier and Pooling

Source: https://github.com/huggingface/pytorch-image-models/blob/main/hfdocs/source/feature_extraction.mdx

Instantiate a model directly without a classifier and global pooling by setting `num_classes=0` and `global_pool=''`. This is an efficient way to get unpooled features from the start.

```python
import torch
import timm
m = timm.create_model('resnet50', pretrained=True, num_classes=0, global_pool='')
o = m(torch.randn(2, 3, 224, 224))
print(f'Unpooled shape: {o.shape}')
```