PyTorch logo PyTorch logo background glow

PyTorch

A python deep learning library based on Torch for building and training neural networks, used for applications in computer vision, natural language processing, and other areas

&

+
Tensor Computation
Provides tensor computation similar to NumPy, with robust GPU acceleration support.
+
Automatic Differentiation
Enables automatic differentiation for creating and training deep neural networks.
+
Performance Mode
Transition seamlessly with TorchScript between eager mode for ease of use and flexibility and graph mode for speed and optimization for production.
+
Distributed Training
Scalable distributed training and performance optimization with support of asynchronous execution of collective operations using the torch.distributed backend.
+
Pre-configured Models
Includes pre-trained models like ResNet, AlexNet, SqueezeNet, VGG, DenseNet, and Inception.
+
Robust Ecosystem
A rich community-built ecosystem of tools and libraries for various domains, from computer vision to reinforcement learning.
+
Scalable
Deploy PyTorch models at scale using TorchServe, which is also environment agnostic allowing deploying models across different cloud platforms.
+
Logging and Metrics
Monitor model performance with built-in logging and metrics features in TorchServe.
+
RESTful Endpoints
Create RESTful endpoints for seamless application integration using TorchServe.
+
Rich Data Loading
Allows to import data from various sources like CSV, image files, and databases, simplifying data pre-processing
+
Automatic Model Optimization
Tools like Quantization Library (TorchQuantization) optimize model size and speed for deployment on resource-constrained devices
+
Visualization Tools
Libraries like TensorBoard integration aid in visualizing model training and performance

Platform

Desktop
Language
PythonC++

Social

System Requirements

#Minimum
1
  • glibc >= v2.17
  • Python >= 3.8
#Minimum
1
macOS 10.15 (Catalina) or above
2
  • Python >= 3.8
#MinimumRecommended
1
  • Windows 7 and greater;
  • Windows Server 2008 r2 and greater
Windows 10 or greater
2
  • Python >= 3.8

Ratings

4.50
5

InfoWorld
4.5
5
based on professional's opinion

Developer

Written in

Python, C++, CUDA, C

Initial Release

24 August 2016

Repository

License

Categories