Beating OpenAI CLIP with 100x less data and compute

Re:Beating OpenAI CLIP with 100x less data and compute by mahnerak on Tuesday February 28, 2023 @02:19PM (#34973569)

I could now find license in the huggingface repo, but it seems like the codebase is Apache 2.0. Are the pretrained weights / checkpoints also covered under this (or other permissive) license?

In other words, can we use it for commercial purposes for free?

[ Parent | on HN ]

Re:Beating OpenAI CLIP with 100x less data and compute by vov_or on Tuesday February 28, 2023 @02:26PM (#34973651)

Hi! Just added Apache2.0 to HF models card. Thanks!

[ Parent | on HN ]

Re:Beating OpenAI CLIP with 100x less data and compute by cosmojg on Tuesday February 28, 2023 @04:19PM (#34975002)

Are the pretraining and training pipelines available anywhere under a FOSS license? I'd love to take a swing at training a mid-fusion model on data other than text and images (e.g., sound, neuron spike trains, etc.)

[ Parent | on HN ]

Re:Beating OpenAI CLIP with 100x less data and compute by grammers on Tuesday February 28, 2023 @02:35PM (#34973762)

Good question, was about to ask the same!

[ Parent | on HN ]

Re:Beating OpenAI CLIP with 100x less data and compute by OkayPhysicist on Tuesday February 28, 2023 @06:37PM (#34976516)

Are weights even copyrightable under US law? It seems like they'd be the output of an automatic process (the training program) the same way the art/text produced by AI models is, which to my understanding makes them not copyrightable material.

[ Parent | on HN ]

Re:Beating OpenAI CLIP with 100x less data and compute by skybrian on Tuesday February 28, 2023 @07:36PM (#34977125)

Compression, even lossy compression, doesn't remove copyright. Whether this is more like compression or a more "transformative use" is something the courts will have to decide someday.

It might be a good time to reread What Color Are Your Bits:

https://ansuz.sooke.bc.ca/entry/23

[ Parent | on HN ]

Re:Beating OpenAI CLIP with 100x less data and compute by renonce on Wednesday March 01, 2023 @03:22AM (#34979990)

There is a lot of manual process involved such as writing training scripts, scarping and processing training data, choosing the best weights among several runs, and spending lots of costly computation. Maybe these should make it copyrightable.

[ Parent | on HN ]