Open victorratts13 opened 5 months ago
Thanks for your continued interest @victorratts13 . As I mentioned in https://github.com/CodeWithKyrian/transformers-php/issues/26, TransformersPHP doesn't support models larger than 2GB for now. OnnxRuntime uses Protobuf buffers to store the models, and they have a 2GB file limit. To make it work, models above that limit have to be split into multiple files, and the current state of the library does not account for downloading and loading multiple files
Rest assured, I have this feature request on top of my mind. Phi 3 is a very fantastic model and gives very impressive results. I'll bump it to the top of the list and work on it as soon as the next release is published (which should be very soon).
Thanks for answering. I'm working on a project responsible for using huggingface AI models with PHP/Laravel. Your library is the most promising I found. If you allow me, I would like to contribute.
Yeah sure. I'm very much open to contribution.
Type of feature request
🌟New Model
Feature description
The
phi3
model is important for certain application scenarios, and its addition would significantly increase the flexibility and usefulness of the library for users. We appreciate your consideration of this request and are available to provide any additional information if needed.Motivation
Currently, when attempting to use the
phi3
model for a task, the following runtime error occurs:This indicates that the
phi3
model is not supported by the library for the given task. However, in response to user requests and the need to utilize thephi3
model, we would like to request the addition of support for this model in the library.Your contribution
Proposal:
phi3
model in functionalities related to the library's pipeline.phi3
model.phi3
model is properly integrated into the existing workflow of the library.