HFT Model Download and Upload options - Uploading and Downloading trained and pre-trained models

[aviraljain99]

This PR is for the following features:

• This will allow for the downloading (and using) of:
  • Models trained using Elpis
  • Pre-trained models from HuggingFace that were fine-tuned or simply used directly
• Uploading
  • Models that were trained using Elpis (and which were downloaded and zipped) can be uploaded.
  • Pre-trained models that were downloaded from HuggingFace (and further fine-tuned or simply used and then downloaded) can also be uploaded.

[benfoley]

Yes, this only supports only HFT engines, will need to be made explicit in the docs.

On 13 Oct 2022, at 10:50, Nicholas Lambourne @.***> wrote:

 @nicklambourne commented on this pull request.

In elpis/endpoints/model.py:

• engine = ENGINES["hft"]
• interface: Interface = app.config["INTERFACE"]
• interface.set_engine(engine) I'm just wondering, if the model isn't deciding what engine it uses, why does the engine need to change here? Is there some inbuilt assumption that only HFT models will be uploaded?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because your review was requested.

[harrykeightley]

@aviraljain99 When you address a comment in the latest commit, can you also resolve the related conversation? Keeps things clean :)