hckrnws
Glad to see that the code[1] is AGPL-3.0 licensed.
Unfortunately, it uses Miniconda, which does not allow usage in companies with more than 200 employees. I think it conflicts with AGPL license. I created a PR to fix that.
Thanks for finding this!
Will get this merged as soon as we can test across platforms.
Is there anything like this but for analyzing thousands of models? Most of this type of tool seem to work best for iterating and experimenting, but I'd actually like to use it for tracking and debugging something in production.
We have a service that runs thousands of custom, user-defined models and when something goes awry having this level of insight would be pretty useful.
Transformer Lab maintainer here. We'd love to build this! Please contact us on our our Discord https://discord.gg/transformerlab or Twitter https://x.com/transformerlab and we'd love to learn more and collaborate.
Seems quite comprehensive. I wonder if there’s a way to “escape” from the system. It looks nice for getting started but in practice at some point you’ll want programmatic access for e.g. advanced API features.
E.g. a well behaving system independent pipeline serialization could work. Didn’t find anything about this with a quick glance of the docs.
Edit: with a bit longer glance there is a plugin system that allows adding custom code quite easily.
One of the maintainers here. Yes this is a big part of our plans. In addition to our plugin system which allows arbitrary Python scripts, we will soon publish how to add decorators to any existing script which can be run externally but logged into Transformer Lab. So you could do training anywhere but trigger evals in the app, for example.
Could this be run as a web interface? For example, getting a GCP instance with GPU and deploying it there and accessing it from the browser.
EDIT: Seems it has a client/server model and the server can be installed somewhere else: https://transformerlab.ai/docs/install/advanced-install
But I'm wondering if this is an electron based app and I could run the client on a web interface as well.
Yes it all works as webapp, in addition to working as an electron app. we haven’t gotten to publishing how in the docs as this feature is brand new.
Short story, run: “npm run start:cloud” in the git repo.
Will make this more known soon.
Perfect, thanks very much! It looks like an amazing tool. Cheers!
Mmm interesting, curious to check it out. The feature list starts with ‘Pick a model’, which I think is the hardest part in the first place. How do you know what mode to pick?
Nit: the website is cut off left and right on mobile (ios)
+1 on android firefox.
how to uninstall?
[flagged]
Crafted by Rajat
Source Code