Commercial use of CodeFormulaV2 and MathWriting training data
#2
by jasperdew - opened
Hi, I'm Jasper, a solo developer working on PDFluent, a PDF SDK with a free desktop editor. Thanks for releasing CodeFormulaV2.
CodeFormulaV2 is CDLA-Permissive-2.0. Its model card says it was trained on SynthFormulaNet and SynthCodeNet. The SynthFormulaNet card says it includes MathWriting rendered at 120 DPI. MathWriting is Google's CC BY-NC-SA 4.0.
- Can a commercial product use and redistribute the CodeFormulaV2 weights under CDLA-Permissive-2.0, given part of the training data is non-commercial?
- Does IBM consider the weights free of the NonCommercial terms of MathWriting?
- Is there a version trained without MathWriting?
A short answer or a pointer to where this is documented would help a lot.
Jasper de Winter, PDFluent