You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
At the moment the Google Drive hosting means there's a manual element to this - downloading the model using a web browser and copying it to the server I'm running it on. In theory there's a way of programmatically accessing the Drive API but it needs a file ID that I don't have access to. Even for browser-based access you need to enable third-party tracking cookies for Drive to download files... (took a wee while to figure out why this wasn't working from Firefox with our standard settings and needed Edge/Chrome)
Would you consider publishing the model somewhere that it can be programmatically reused, and accompanied by a model card?
Is this something the people maintaining the Turing sci.vision site are able to support you, as a researcher, to do? If sci.vision isn't the natural home for this kind of project, it would be good to know why!
If that's not an option, would it make sense to upload them to the HuggingFace model hub? With the caveat that I've never tried this, only consumed their services. I'd be up for setting up a call to step through trying it, though.
Let me know who it makes sense to tag into this discussion from the scivision angle...
The text was updated successfully, but these errors were encountered:
Hello! Thank you for making all this available for re-use.
I'm experimenting with using the model weights available in Google Drive to process some freshwater plankton data
At the moment the Google Drive hosting means there's a manual element to this - downloading the model using a web browser and copying it to the server I'm running it on. In theory there's a way of programmatically accessing the Drive API but it needs a file ID that I don't have access to. Even for browser-based access you need to enable third-party tracking cookies for Drive to download files... (took a wee while to figure out why this wasn't working from Firefox with our standard settings and needed Edge/Chrome)
Would you consider publishing the model somewhere that it can be programmatically reused, and accompanied by a model card?
Is this something the people maintaining the Turing sci.vision site are able to support you, as a researcher, to do? If sci.vision isn't the natural home for this kind of project, it would be good to know why!
If that's not an option, would it make sense to upload them to the HuggingFace model hub? With the caveat that I've never tried this, only consumed their services. I'd be up for setting up a call to step through trying it, though.
Let me know who it makes sense to tag into this discussion from the scivision angle...
The text was updated successfully, but these errors were encountered: