ofa-sys / instag Goto Github PK

Thanks for your awesome work. When I download the model from huggingface and use fastchat to deploy it, the result has the right json format, but the tag result is not reasonable. However, when I input the same query to the online demo, the tag result is as expected.

I use the system prompt of vicua_v1.1 and do not use any demonstration. Is there anything I miss?

Thanks for your excellent work.
I would really like to follow your work but the lack of data and tags makes it hard to do so. Thus I wonder whether you will release the data and corresponding tags in the near future?
I believe it would be very influential work！

Hello!

The link to the ModelScope Demo seems to be broken. It redirects to just the home page.
Would love to try out the demo if possible!

Thanks

Hi, your paper has already been accepted by ICLR 2024 as a poster. You have mentioned in OpenReview We will release the ChatGPT-tagged data pool and local tagger to promote the development of more precise and cheap tag annotators.

I am wondering when the dataset will be available. Many thanks in advance!

Hi authors,

Thanks for the great work! Got couple of questions after reading your paper:

(1) Do you plan to release the code for tag normalization, e.g. frequency filtering, rule & semantic aggregation, association aggregation?

(2) When you trained the InsTagger model, did you use the normalized tags or raw tags?

Thanks!

Sorry, I'd like to use your tool to tag my dataset, but I seem to have not found guidance on the parameter settings for SamplingParams。