Enjoy Hugging Face's model library with Valohai's templates
Henrik SkogströmWe're launching the Valohai Ecosystem Library today – a collection of ready-made templates and connectors you can use across your Valohai projects.
At launch, the Ecosystem Library will have three database connectors, making it easy to pull data from common databases and a collection of templates for Hugging Face Transformers, making it easy to experiment with the largest open-source model library.
Since we've previously written about the database connectors, let's drill into the Hugging Face templates.
Language models are progressing at a breakneck speed, and Hugging Face is the hub for the latest and greatest in open-source ML. We've built a set of Hugging Face templates that make it super simple to fine-tune any Transformers models for a specific task and do batch inference – without writing your own code.
Here's a short introduction to how these Hugging Face templates work:
We've included templates for three different tasks:
Task | Example Model |
---|---|
Text Classification | distilbert-base-uncased |
Summarization | sshleifer/distilbart-xsum-12-1 |
Question Answering | distilbert-base-uncased |
Each task has templates for fine-tuning & batch inference and example datasets. These templates can be used with any Transformers models from the Hugging Face model library – just copy the repository name from Hugging Face and paste it into Valohai.
All connectors and templates in the Valohai Ecosystem are completely open-source, so you can easily fork and repurpose them. You can find them on GitHub.
The Valohai Ecosystem is now available for all Valohai users. If you are new to Valohai and looking to get started, book a demo today!