# SimCLR

**Repository Path**: xiaoxu85/SimCLR

## Basic Information

- **Project Name**: SimCLR
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-01-05
- **Last Updated**: 2021-01-05

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# SimCLR
PyTorch implementation of SimCLR: A Simple Framework for Contrastive Learning of Visual Representations by T. Chen et al.
Including support for:
- Distributed data parallel training
- Global batch normalization
- LARS (Layer-wise Adaptive Rate Scaling) optimizer.

[Link to paper](https://arxiv.org/pdf/2002.05709.pdf)

Open SimCLR in Google Colab Notebook (with TPU support)

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1ObAYvVKQjMG5nd2wIno7j2y_X91E9IrX)

Open SimCLR results comparison on tensorboard.dev:

<p align="left">
  <a href="https://tensorboard.dev/experiment/A3laNdafRBes0oR45Y6LiA/#scalars" target="_blank">
    <img src="https://github.com/Spijkervet/SimCLR/blob/master/media/tensorboard.png?raw=true" height="40"/>
  </a>
</p>

### Quickstart (fine-tune linear classifier)
This downloads a pre-trained model and trains the linear classifier, which should receive an accuracy of ±`82.9%` on the STL-10 test set.
```
git clone https://github.com/spijkervet/SimCLR.git && cd SimCLR
wget https://github.com/Spijkervet/SimCLR/releases/download/1.2/checkpoint_100.tar
sh setup.sh || python3 -m pip install -r requirements.txt || exit 1
conda activate simclr
python -m testing.logistic_regression --dataset=STL10 --model_path=. --epoch_num=100 --resnet resnet50
```

#### CPU
```
wget https://github.com/Spijkervet/SimCLR/releases/download/1.1/checkpoint_100.tar -O checkpoint_100.tar
python -m testing.logistic_regression --model_path=. --epoch_num=100 --resnet=resnet18 --logistic_batch_size=32
```

### Quickstart (pre-train ResNet encoder using SimCLR)
```
python main.py
```
With distributed data parallel (DDP) training:
```
CUDA_VISIBLE_DEVICES=0 python main.py --nodes 2 --nr 0
CUDA_VISIBLE_DEVICES=1 python main.py --nodes 2 --nr 1
CUDA_VISIBLE_DEVICES=2 python main.py --nodes 2 --nr 2
CUDA_VISIBLE_DEVICES=N python main.py --nodes 2 --nr 3
```


### Results
These are the top-1 accuracy of linear classifiers trained on the (frozen) representations learned by SimCLR:

| Method  | Batch Size | ResNet | Projection output dimensionality | Epochs | Optimizer | STL-10 | CIFAR-10
| ------------- | ------------- | ------------- | ------------- | ------------- | ------------- | ------------- | ------------- |
| SimCLR + Linear eval. | 256 | ResNet50 | 64 | 100 | Adam | **0.829** | **0.833** | 
| SimCLR + Linear eval. | 256 | ResNet50 | 64 | 100 | LARS | 0.783 | - | 
| SimCLR + Linear eval. | 256 | ResNet18 | 64 | 100 |  Adam | 0.765  | - |
| SimCLR + Linear eval. | 256 | ResNet18 | 64 | 40 | Adam | 0.719  | - |
| SimCLR + Linear eval. | 512 | ResNet18 | 64 | 40 | Adam | 0.71 | - |
| Logistic Regression | - | - | - | 40 | Adam | 0.358 | 0.389 |



### Pre-trained models
| ResNet (batch_size, epochs) | Optimizer | STL-10 Top-1 |
| ------------- | ------------- | ------------- |
| [ResNet50 (256, 100)](https://github.com/Spijkervet/SimCLR/releases/download/1.2/checkpoint_100.tar) | Adam | **0.829** |
| [ResNet18 (256, 100)](https://github.com/Spijkervet/SimCLR/releases/download/1.1/checkpoint_100.tar) | Adam | 0.765 |
| [ResNet18 (256, 40)](https://github.com/Spijkervet/SimCLR/releases/download/1.0/checkpoint_40.tar) | Adam | 0.719 |

`python -m testing.logistic_regression --model_path=. --epoch_num=100`

#### LARS optimizer
The LARS optimizer is implemented in `modules/lars.py`. It can be activated by adjusting the `config/config.yaml` optimizer setting to: `optimizer: "LARS"`. It is still experimental and has not been thoroughly tested.

## What is SimCLR?
SimCLR is a "simple framework for contrastive learning of visual representations". The contrastive prediction task is defined on pairs of augmented examples, resulting in 2N examples per minibatch. Two augmented versions of an image are considered as a correlated, "positive" pair (x_i and x_j). The remaining 2(N - 1) augmented examples are considered negative examples. The contrastive prediction task aims to identify x_j in the set of negative examples for a given x_i.

<p align="center">
  <img src="https://github.com/Spijkervet/SimCLR/blob/master/media/architecture.png?raw=true" width="500"/>
</p>

## Usage
Run the following command to setup a conda environment:
```
sh setup.sh
conda activate simclr
```

Or alternatively with pip:
```
pip install -r requirements.txt
```

Then, simply run for single GPU or CPU training:
```
python main.py
```

For distributed training (DDP), use for every process in nodes, in which N is the GPU number you would like to dedicate the process to:
```
CUDA_VISIBLE_DEVICES=0 python main.py --nodes 2 --nr 0
CUDA_VISIBLE_DEVICES=1 python main.py --nodes 2 --nr 1
CUDA_VISIBLE_DEVICES=2 python main.py --nodes 2 --nr 2
CUDA_VISIBLE_DEVICES=N python main.py --nodes 2 --nr 3
```

`--nr` corresponds to the process number of the N nodes we make available for training.

### Testing
To test a trained model, make sure to set the `model_path` variable in the `config/config.yaml` to the log ID of the training (e.g. `logs/0`).
Set the `epoch_num` to the epoch number you want to load the checkpoints from (e.g. `40`).

```
python -m testing.logistic_regression
```

or in place:
```
python -m testing.logistic_regression --model_path=./save --epoch_num=40
```


## Configuration
The configuration of training can be found in: `config/config.yaml`. I personally prefer to use files instead of long strings of arguments when configuring a run. An example `config.yaml` file:
```
# train options
batch_size: 256
workers: 16
start_epoch: 0
epochs: 40
dataset_dir: "./datasets"

# model options
resnet: "resnet18"
normalize: True
projection_dim: 64

# loss options
temperature: 0.5

# reload options
model_path: "logs/0" # set to the directory containing `checkpoint_##.tar` 
epoch_num: 40 # set to checkpoint number

# logistic regression options
logistic_batch_size: 256
logistic_epochs: 100
```

## Logging and TensorBoard
To view results in TensorBoard, run:
```
tensorboard --logdir runs
```

## Optimizers and learning rate schedule
This implementation features the Adam optimizer and the LARS optimizer, with the option to decay the learning rate using a cosine decay schedule. The optimizer and weight decay can be configured in the `config/config.yaml` file.
<p align="center">
  <img src="https://github.com/Spijkervet/SimCLR/blob/master/media/lr_cosine_decay_schedule.png?raw=true" width="400"/>
</p>

#### Dependencies
```
torch
torchvision
tensorboard
pyyaml
```