Skip to content
This repository has been archived by the owner on Nov 21, 2022. It is now read-only.

Latest commit

聽

History

History
82 lines (48 loc) 路 3.59 KB

CHANGELOG.md

File metadata and controls

82 lines (48 loc) 路 3.59 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.2.5] - 2022-11-21

Fixed

  • Fixed loading HF model (#306)
  • Fixed passing config name to CNNDailyMailSummarizationDataModule (#310)
  • Fixed move pipeline to self.device as default (#309)

[0.2.4] - 2022-11-03

Changed

  • Added support for Lightning v1.8.0 (#297)

[0.2.3] - 2022-10-08

Changed

  • Use lightning-utilities for compatibility (#292)

[0.2.2] - 2022-10-07

Changed

  • Added support for Lightning v1.7.0 (#284)
  • Allow generation, fix examples (#271)

Fixed

  • Fixed LightningCLI compatibility (#288)

[0.2.1] - 2022-06-28

Changed

  • Simplified Large Model Support/Add Large Model Training (#269)
  • Refactored the code for model creation (#268)

[0.2.0] - 2022-06-23

Added

  • Added big model support (#263)
  • Added support for Trainer.predict method (#261)
  • Added ViT Image Classification Support (#252)
  • Allow streaming datasets for the language modeling task (#256)
  • Micro to macro conversion for Classification Metrics (#255)
  • Added ability to set kwargs for pipeline in load_from_checkpoint (#204)
  • Added Masked Language Modeling (#173)
  • Added more schedulers (#143)

Changed

  • Refactored for new class based approach (#243)
  • Flatten inheritance (#245)
  • Removed configs (#264)
  • Removed config/hydra from repo (#262)
  • Changed gpus to devices for compatability (#239)

Fixed

  • Fixed Token Classification (#265)
  • Fixed broken bleu metrics (#228)
  • Fixed a bug preventing a model evaluating on the test split (#186)
  • Reflect default settings changed to using ddp (#175)

[0.1.0] - 2022-04-21

  • EVERYTHING