Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Release 1.0.1 #133

Merged
merged 101 commits into from
Nov 13, 2024
Merged

Release 1.0.1 #133

merged 101 commits into from
Nov 13, 2024

Conversation

kasnerz
Copy link
Collaborator

@kasnerz kasnerz commented Oct 29, 2024

Version 1.0.1

First official release 🎉

Changes compared to the pre-release version:

Pip package

We release factgenie as a PyPI package to make its installation easier.

(Note that the recommended way to install factgenie is still as an editable Python package. That setup enables to write custom dataset classes.)

Resolves #3 .

File structure

We reorganized the file structure to make it more intuitive and sustainable. See #120 for more details.

Resolves #120 .

Multiple annotations within a campaign

We enable multiple annotations within a campaign, both during data collection and in the browsing interface.

Resolves #60.

Support for VLLM

We support VLLM as a backend for running LLMs, including constrained decoding for the annotations.

Resolves #75.

Prompt and instruction wizards

We support pre-filling prompts for LLMs and instructions for the crowdworkers based on the annotation span categories.

Resolves #108 .

Better reproducibility

We save more detailed information in the generated outputs to enhance reproducibility.

Resolves #130 .

Minor fixes

We implemented many other minor fixes.

Resolves #111, resolves #144, resolves #150.

Documentation and testing

We completely updated the documentation on wiki and tested factgenie in various setups.

Resolves #50.

@kasnerz kasnerz changed the title Release 1.0.0 WIP: Release 1.0.0 Oct 29, 2024
oplatek and others added 22 commits November 6, 2024 15:16
CLI tests using factgenie list datasets and factgenie list campaigns
fix parse_crowdsourcing_config default empty strings
fix factgenie launch; add ollama service to docker-compose
* add VLLMetric and OpenAIMetric supporting structured decoding

* add example for VLLM inference
Completely changed parsing LLM-eval annotations
Now we use pydantic and enforce structure

* rename type to annotation_type for parsing LLM outputs and in LLM prompts
and new functionality factgenie download -d DATASET_ID
@kasnerz kasnerz changed the title WIP: Release 1.0.0 Release 1.0.0 Nov 13, 2024
@kasnerz kasnerz changed the title Release 1.0.0 Release 1.0.1 Nov 13, 2024
@kasnerz kasnerz merged commit 3fe1a36 into main Nov 13, 2024
4 checks passed
@oplatek oplatek deleted the release-1.0.0 branch November 13, 2024 14:59
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment