Hello! During the convolutional neural network and sequence model course, in the videos that Andrew developed, he talks about models with permissions to be used or models with parameters already trained or training sets to use the models. I would be very grateful if you could give me links or indicate portals that allow me to work with these models and data. Also, are there models that are already implemented and can be used for commercial purposes?
How do licenses work in DeepLearning models? Are there open source licenses? Can the models published by those who have conducted the research be used to develop applications? Is it possible to modify and improve them. How does this work in research to respect copyright and how do I know what permissions a model has?
There are no licenses required for the DLAI courses.
All of the models you need are provided in the programing assignment labs.
I am sorry @TMosh. I don’t seem to explain myself very well. My query is regarding Deeplearning models that I can find within the course or outside the course. Could you recommend a web portal or links to algorithms or models that can be used for commercial purposes, for example the LLM LLama2. Do you know of any website that brings together the main models for computer vision and language models?
No, I don’t know of such a comprehensive list.
Ok, thank you @TMosh
Thank you very much! @balaji.ambresh. Your help is very important for those of us who are just starting out in this. I would like to know how this world works between open source, open science and private or closed AI developments. I am very grateful for your contribution.
Balaji has given us a much wider ranging set of links, but it’s also worth noting that any of the standard ML/DL platforms also provides a library of pretrained models that you can use directly or for Transfer Learning. E.g. in most of the courses here we use TensorFlow/Keras. Here’s the top level documentation page for the pretrained models provided by Keras.
You’re welcome.
Each model has a model card (example) or is tagged with license information as shown below:
If there isn’t one, look inside README.md (example).
Do reach the contributor for further information if no usage details are specified. You can upload custom models as well (see docs).
As far as closed source development is concerned, consider using a private repository for storing the model. Git LFS is one way to manage large model files. See mlops specialization for details on making models production ready.
Thank you very much @paulinpaloalto, this is a very good summary for the Keras framework. This is also very useful. Thanks for your help!
Thank you very much again. It is clearer to me after the explanation.
Best regards!