A fast & nice-looking PyTorch library for geometric deep learning (NN on graphs and other irregular structures).
— Yann LeCun (@ylecun) March 8, 2019
Code: https://t.co/OoOaNy9naF
Paper: https://t.co/CIzrCTqtHg
"Fast Graph Representation... https://t.co/OoOaNy9naF
PyTorch Geometric has been growing into a fantastic library for graph neural nets and related methods. Great work by @rusty1s and Jan Eric Jenssen- https://t.co/HTMf4QuyZO https://t.co/3LVqiXD5KJ
— Thomas Kipf (@thomaskipf) March 7, 2019
Pytorch-bert v0.6 is out with OpenAI's pretrained GPT-2 🦄 small model & the usual accompanying example scripts to use it.
— Thomas Wolf (@Thom_Wolf) February 18, 2019
Now... can you guys wait that the ACL deadline has passed to release any crazy new transformer? 😅
Thanks, you are the best! 🥰 pic.twitter.com/Xn06E4fM4c
Neural Pipeline – NN training pipeline based on #PyTorch to accelerate experiments https://t.co/k2YMKLoGnu
— Alexandr Kalinin (@alxndrkalinin) February 14, 2019
- flexible training
- checkpoint management
- metric visualization
- best practices (e.g. learning rate decaying & hard negative mining)
- DVC compatible metric logging pic.twitter.com/IcGE7rePdh
PT-BERT 0.5 out💥
— Thomas Wolf (@Thom_Wolf) February 11, 2019
Pretty big release w. not 1 but TWO new pretrained models:
-classic: OpenAI's GPT
-brand-new: Transformer-XL by Google/CMU
As always both should be super easy to use
So...BERT now stands for Big-&-Extending-Repository-of-Transformers😅
Happy Transfer Learning! pic.twitter.com/yIdTHHJKKA
PyTorch implementation of "Super-Realtime Facial Landmark Detection and Shape Fitting by Deep Regression of Shape Model Parameters" https://t.co/kAFzsIElJx #deeplearning #machinelearning #ml #ai #neuralnetworks #datascience #pytorch
— PyTorch Best Practices (@PyTorchPractice) February 10, 2019
Very useful package for PyTorch: "torchbearer: A model fitting library for PyTorch" https://t.co/kSSGlJw7dM pic.twitter.com/08ms6MvHQt
— Learn PyTorch (@learnpytorch) February 6, 2019
[v1.0.1] A purely bug fix release.
— PyTorch (@PyTorch) February 5, 2019
Note, our Conda install commands have changed.
Feature identifiers such as `cuda100` have changed to `cudatoolkit=10.0`.
See the new commands and get the latest version from https://t.co/DeaBDSRxs8
Read more at: https://t.co/lRmkO5ZfOT
PyTorch Implementation of XNOR-Net https://t.co/NyQ95DvPtR #deeplearning #machinelearning #ml #ai #neuralnetworks #datascience #pytorch
— PyTorch Best Practices (@PyTorchPractice) February 3, 2019
Pytorch implementation of RA-retrofit https://t.co/iwaBgQs4tZ #deeplearning #machinelearning #ml #ai #neuralnetworks #datascience #pytorch
— PyTorch Best Practices (@PyTorchPractice) January 27, 2019
This has been my experience with the latest TF versions too. A common mistake of ignoring complexity cost in engineering and optimizing for the wrong things (the “gmail redesign effect”). Wish I could switch to PyTorch but some production use cases are not covered by it yet.. pic.twitter.com/s7SKX73kuc
— Denny Britz (@dennybritz) January 26, 2019
PyTorch is now installed by default on Google Colaboratory.
— PyTorch (@PyTorch) January 25, 2019
Go ahead and `import torch`, `import torchvision`, `import torchtext` pic.twitter.com/PISPoegATV
