More information the model, chat template and examples.
Hi there,
Congratulations on open-sourcing such a brilliant model. Hinglish LLMs are definitely something the community can benefit quite a lot from.
Could you please provide more information about the model, chat template, how to use it and some examples in the model card?
I believe this would result in a considerable increase in the adoption of the model.
Lastly, are there plans to train smaller 7B/ 13B scale models as well?
Cheers!
VB
Yes, we have made a demo video for this: https://www.linkedin.com/feed/update/urn:li:activity:7169266339990233088
Chat template is same as Llama-70b: [INST] query[/INST] answer
Stop words are: [INST]
I will add it to model card.
We have made llama-7b (closed source), llama-13b (open source - 3 variants - Eng to Hindi, Hindi to Eng and Hindi chat) and Mixtral(closed source) in the past. Now, working on other models (particularly Deepseek-7b-Maths).
Our previous models were shared on our LinkedIn:
https://www.linkedin.com/feed/update/urn:li:activity:7143255593103810560
https://www.linkedin.com/feed/update/urn:li:activity:7147075414304112640
And can be downloaded from Github links on those posts.