GitHub · Where software is built

This repository was archived by the owner on Apr 23, 2025. It is now read-only.

Labels Milestones

Checkpointing should be extended to optimizer and training loop state

#726

· BradLarson opened

on Dec 28, 2020

Complex Loss Functions Inside TrainingLoop

#720

· xanderdunn opened

on Dec 12, 2020

Better error message on unzip failure

good first issue

#675

· texasmichelle opened

on Oct 1, 2020

GPT-2 currently exhausts all available GPU memory on an 8 GB GPU

#673

· BradLarson opened

on Sep 29, 2020

Add logging callbacks for training statistics

good first issue

#663

· BradLarson opened

on Sep 2, 2020

Add documentation to the Gym models to specify additional dependencies for DQN

good first issue

#657

· BradLarson opened

on Aug 10, 2020

Implementing graph convolutional neural networks

#643

· hector-crean opened

on Jul 20, 2020

Transformer model requires more parameters than supported on TPU

#638

· BradLarson opened

on Jul 16, 2020

[WordSeg] Use <code>Character</code> instead of <code>String</code> in <code>Alphabet</code>

#600

· texasmichelle opened

on Jun 10, 2020

[WordSeg] Alphabet can only contain characters from training

#599

· texasmichelle opened

on Jun 10, 2020

Find a more descriptive word for <code>numericalized</code>

good first issue

#598

· texasmichelle opened

on Jun 10, 2020

Add a DistilGPT2 variant

#549

· texasmichelle opened

on May 19, 2020