Post
1319
Awesome new project by
numind
Multi-modal structured data extraction. Give it a JSON schema, some sort of data (image, text) and optionally examples, and get the filled-in json schema as a result.
Kind of crazy.
THOUGHTS BELOW:
Its super fast and super effective. It is so practical.
Im getting basically instant results on an A100(40gb) with VLLM and the 4b size.
so exciting. AND COMMERCIAL FREIDNLYYYY
numind
Multi-modal structured data extraction. Give it a JSON schema, some sort of data (image, text) and optionally examples, and get the filled-in json schema as a result.
Kind of crazy.
THOUGHTS BELOW:
Its super fast and super effective. It is so practical.
Im getting basically instant results on an A100(40gb) with VLLM and the 4b size.
so exciting. AND COMMERCIAL FREIDNLYYYY