PyTorch implementation of SimSiam https//arxiv.org/abs/2011.10566.
facebookresearch/simsiam is ranked 11,026 with 1,236 stars.
Same-category comparison
facebookresearch/simsiam is ranked 11,026 among these 8 peers; ivan-vasilev/neuralnetworks is ranked 11,027.
Same-category comparison| Name | Rank | Metrics |
|---|
| atpaino/deep-text-corrector | 11,018 | - Language:
- Python
- Stars:
- 1,237
- Forks:
- 258
- Description:
- Deep learning models trained to correct input errors in short, message-like text.
|
|---|
| openseg-group/openseg.pytorch | 11,020 | - Language:
- Python
- Stars:
- 1,237
- Forks:
- 141
- Description:
- The official Pytorch implementation of OCNet, OCRNet, and SegFix.
|
|---|
| AGI-Edgerunners/LLM-Adapters | 11,022 | - Language:
- Python
- Stars:
- 1,236
- Forks:
- 117
- Description:
- Code for our EMNLP 2023 Paper: "LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models".
|
|---|
| SciML/NeuralPDE.jl | 11,024 | - Language:
- Julia
- Stars:
- 1,236
- Forks:
- 252
- Description:
- Physics-Informed Neural Networks (PINN) Solvers of (Partial) Differential Equations for Scientific Machine Learning (SciML) accelerated simulation.
|
|---|
| ivan-vasilev/neuralnetworks | 11,027 | - Language:
- Java
- Stars:
- 1,236
- Forks:
- 345
- Description:
- java deep learning algorithms and deep neural networks with gpu acceleration.
|
|---|
| BR-IDL/PaddleViT | 11,029 | - Language:
- Python
- Stars:
- 1,235
- Forks:
- 328
- Description:
- :robot: PaddleViT: State-of-the-art Visual Transformer and MLP Models for PaddlePaddle 2.0+.
|
|---|
| jakesnell/prototypical-networks | 11,031 | - Language:
- Python
- Stars:
- 1,235
- Forks:
- 270
- Description:
- Code for the NeurIPS 2017 Paper "Prototypical Networks for Few-shot Learning".
|
|---|
| jfzhang95/pytorch-video-recognition | 11,038 | - Language:
- Python
- Stars:
- 1,234
- Forks:
- 251
- Description:
- PyTorch implemented C3D, R3D, R2Plus1D models for video activity recognition.
|
|---|
Key parameters and specifications
Key parameters and specifications| Parameter | Value |
|---|
| Language | Python |
|---|
| Category | Model Development |
|---|
| Subcategory | Architecture |
|---|
| Stars | 1,236 |
|---|
| Forks | 174 |
|---|
| Contributors | 0 |
|---|
| Created date | 2021-03-29 |
|---|
| Updated date | 2026-09-23 |
|---|
| Stars in 1 day | 0 |
|---|
| Stars in 7 days | 0 |
|---|
| Rank | 11,026 |
|---|
| Previous rank | 10,983 |
|---|
| Rank change | -43 |
|---|
| 7-day stars rank | 12,856 |
|---|
| Previous 7-day stars rank | 12,875 |
|---|
| 7-day stars rank change | 19 |
|---|
| Total stars rank | 9,467 |
|---|
| Previous total stars rank | 9,457 |
|---|
| Total stars rank change | -10 |
|---|
| Country | USA |
|---|
README summary
SimSiam: Exploring Simple Siamese Representation Learning This is a PyTorch implementation of the SimSiam paper: Preparation Install PyTorch and download the ImageNet dataset following the official PyTorch ImageNet training code. Similar to MoCo, the code release contains minimal modifications for both unsupervised pre-training and linear classification to that code. In addition, install apex for the LARS implementation needed for linear classification. Unsupervised Pre-Training Only **multi-gpu**, **DistributedDataParallel** training is supported; single-gpu or DataParallel training is not supported
. To do unsupervised pre-training of a ResNet-50 model on ImageNet in an 8-gpu machine, run: The script uses all the default hyper-parameters as described in the paper, and uses the default augmentation recipe from MoCo v2. The above command performs pre-training with a non-decaying predictor learning rate for 100 epochs, corresponding to the last row of Table 1 in the paper. Linear Classification With a pre-trained model, to train a supervised linear classifier
on frozen features/weights in an 8-gpu machine, run: The above command uses LARS optimizer and a default batch size of 4096. Models and Logs Our pre-trained ResNet-50 models and logs: pre-train epochs batch size pre-train ckpt pre-train log linear cls. 3 Settings for the above: 8 NVIDIA V100 GPUs, CUDA 10. Transferring to Object Detection Same as MoCo for object detection transfer, please see moco/detection. License This project is under the CC
-BY-NC 4.
Frequently asked questions
How many stars and forks are recorded?
facebookresearch/simsiam has 1,236 stars and 174 forks.
What is the repository rank?
facebookresearch/simsiam is ranked 11,026.
What language and category are recorded?
facebookresearch/simsiam has 1,236 stars, uses Python and belongs to Model Development.
Which repository is closest in rank?
facebookresearch/simsiam is compared with ivan-vasilev/neuralnetworks, ranked 11,027 with 1,236 stars.
SimSiam: Exploring Simple Siamese Representation Learning
This is a PyTorch implementation of the SimSiam paper:
@Article{chen2020simsiam,
author = {Xinlei Chen and Kaiming He},
title = {Exploring Simple Siamese Representation Learning},
journal = {arXiv preprint arXiv:2011.10566},
year = {2020},
}
Preparation
Install PyTorch and download the ImageNet dataset following the official PyTorch ImageNet training code. Similar to MoCo, the code release contains minimal modifications for both unsupervised pre-training and linear classification to that code.
In addition, install apex for the LARS implementation needed for linear classification.
Unsupervised Pre-Training
Only multi-gpu, DistributedDataParallel training is supported; single-gpu or DataParallel training is not supported.
To do unsupervised pre-training of a ResNet-50 model on ImageNet in an 8-gpu machine, run:
python main_simsiam.py \
-a resnet50 \
--dist-url 'tcp://localhost:10001' --multiprocessing-distributed --world-size 1 --rank 0 \
--fix-pred-lr \
[your imagenet-folder with train and val folders]
The script uses all the default hyper-parameters as described in the paper, and uses the default augmentation recipe from MoCo v2.
The above command performs pre-training with a non-decaying predictor learning rate for 100 epochs, corresponding to the last row of Table 1 in the paper.
Linear Classification
With a pre-trained model, to train a supervised linear classifier on frozen features/weights in an 8-gpu machine, run:
python main_lincls.py \
-a resnet50 \
--dist-url 'tcp://localhost:10001' --multiprocessing-distributed --world-size 1 --rank 0 \
--pretrained [your checkpoint path]/checkpoint_0099.pth.tar \
--lars \
[your imagenet-folder with train and val folders]
The above command uses LARS optimizer and a default batch size of 4096.
Models and Logs
Our pre-trained ResNet-50 models and logs:
pre-train
epochs
batch
size
pre-train
ckpt
pre-train
log
linear cls.
ckpt
linear cls.
log
top-1 acc.
100
512
link
link
link
link
68.1
100
256
link
link
link
link
68.3
Settings for the above: 8 NVIDIA V100 GPUs, CUDA 10.1/CuDNN 7.6.5, PyTorch 1.7.0.
Transferring to Object Detection
Same as MoCo for object detection transfer, please see moco/detection.
License
This project is under the CC-BY-NC 4.0 license. See LICENSE for details.