Redistribution licence and origin of the ONNX conversions
Hello. We are reviewing the redistribution terms for a Windows desktop app that performs inference locally. We are considering bundling these ONNX conversions without changing their weight files:
- Xenova/clip-vit-base-patch32, revision d15189d7028b43f1d3e65039190477f6af591c2a
- Xenova/trocr-small-handwritten, revision 2432e24d184b1d964d07ed04f5d9e21d31a59141
- Xenova/vit-gpt2-image-captioning, revision 215b4edcb7ec1fad5905a18a03f7b2007f6fabd0
Their cards identify the base models but do not state an independent licence. Do the applicable base-model permissions and notices also cover these ONNX conversion contributions? Are there additional redistribution notices or restrictions, and are the exact base-model revisions recorded anywhere? We are not requesting a purchase or entering a commercial agreement. Thank you.
Hello, a brief follow-up to our question above. Could you please clarify whether the applicable upstream licences also cover your ONNX conversion contributions for the three listed revisions, and identify any additional notices or redistribution restrictions? A link to the applicable licence and exact upstream revisions would be very helpful. We are checking redistribution with a proprietary Windows app performing inference locally; this is an information request, not acceptance of a purchase or contract. Thank you for your time.
thanks! i have updated this repo to be MIT (based on https://github.com/openai/CLIP/blob/main/LICENSE)
Xenova/trocr-small-handwritten is also MIT (based on https://github.com/microsoft/unilm/blob/master/LICENSE)
Xenova/vit-gpt2-image-captioning is Apache-2.0 (based on https://huggingface.co/nlpconnect/vit-gpt2-image-captioning)