Spaces:

argilla

/

synthetic-data-generator

like 314

Running

Hi @thomwolf ,

Thanks for the feedback.

• I agree that would be nice. @sdiazlor is working on a slight redesign and improvement of the task.
• FYI @sdiazlor is also working on addingevaluation/judging to the space. (this task will use the proposed space next to the input boxes to preview the dataset to evaluate, hence we had not moved them there at first.
• Adding custom generation model handles is nice but somewhat complex due to the fact there is no clear selection of models that are available as serverless inference endpoints with a high enough rate limit (70B is lower than 7B and might not scale). Additionally, we need a specific query template-logic to work with Magpie generation. We can for sure add an easy dropdown to add some Qwen an other llama models because these have already been implemented.
• After this redesign, I will for sure add a more robust choice of models and stores because I want to push this through GH and a pip install as a more formal "package" to easier get traction through socials and SEO.

@thomwolf , we just pushed a new and cleaner version :)

@thomwolf also shipped a basic feature to swap uit models (not UI-based at the moment) but it might help it go more viral, since we rely on the OpenAI endpoint specs, hence people can use a lot more models. https://github.com/argilla-io/synthetic-data-generator

Nice, taking a look!

Very nice! Small feedback:
- when not clicked on "logged in" in the error is a bit cryptic: ""Failed to get organizations: Invalid user token. If you didn't pass a user token, make sure you are properly logged in by executing huggingface-cli login, and if you did pass a user token, double-check it's correct.. See if you are logged and connected: https://huggingface.co./settings/connected-applications."

@thomwolf I will review the message. It is supposed to disable the UI when not logged in, but this behaviour seems to differ per browser (settings) and user. Still trying to fix it.

@thomwolf resolved. Thanks for the feedback. We are launching today. Hope we can add eval and RAG next.