Skip to content

v0.3.0
Compare
Choose a tag to compare
@yifanmai yifanmai released this 01 Nov 22:01
· 886 commits to main since this release
v0.3.0
8ea285f
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

Models

  • Added support for Lit-GPT (#1792)
  • Added support for stop sequences in HuggingFaceClient (#1892, #1909)
  • Added Mistral 7B model (#1906)
  • Added IDEFICS model (#1871)
  • Added Anthropic Claude 2 (#1900)

Scenarios

  • Added 31 scenarios from CLEVA for evaluation of Chinese language models (#1824, #1864)
  • Added VQA scenario model (#1871)
  • Adddd support for running MCQA scenarios from users' JSONL files (#1889)

Metrics

  • Fixed a bug that prevented using Anthropic Claude for model critique (#1862)

Frontend

Framework

  • Added support for multi-modal scenarios and Vision Language Model (VLM) evaluation (#1871)
  • Added support for Python 3.9 and 3.10 (#1897)
  • Added a new Tokenizer class in preparation for removing tokenize() and decode() from Client in a future release (#1874)
  • Made more dependencies optional instead of required, and added install command suggestions (#1834, #1961)
  • Added support for configuring users' model deployments through YAML configuration files (#1861)

Evaluation Results

  • Added evaluation results for Stanford Alpaca, MosaicML MPT, TII UAE Falcon, LMSYS Vicuna

Contributors

Thank you to the following contributors for your work on this HELM release!

Contributors

yifanmai, percyliang, and 16 other contributors
Assets 2
Loading