Review
Of all the TinyLlama models that I downloaded to date (September 1, 2024), I could say that it was the best model that I was lucky enough to download for my cell phone, even surpassing the QWEN 2 in several aspects.
The testing has been carried out in Spanish, in which the level of responses is surprising and the great work and effort put in by the development team in this work is evident, even in the presentation of the model where it is precisely detailed what data to put in for its correct operation, which almost no one does.
It's a shame that its development has stopped.
Thanks.
Of all the TinyLlama models that I downloaded to date (September 1, 2024), I could say that it was the best model that I was lucky enough to download for my cell phone, even surpassing the QWEN 2 in several aspects.
The testing has been carried out in Spanish, in which the level of responses is surprising and the great work and effort put in by the development team in this work is evident, even in the presentation of the model where it is precisely detailed what data to put in for its correct operation, which almost no one does.
It's a shame that its development has stopped.
Thanks.
Hello Novaciano,
Thanks a lot for the review! We are happy to hear that it met your requirements, even in Spanish. As we are a Pakistani Company, we could not test it out in Spanish ourselves, which meant that we could only perform the quality check in Urdu, English and German, and it really makes us happy to hear that it also performs well in Spanish.
We are also very happy that you red trough the model card, which includes all information to make to model run as well as possible.
As for the development, as some may have already noticed, we are very sorry that it stopped. We are currently out of money, but we are doing our best to keep the service up. We are currently developing a closed-source flagship model, but since we heard that you liked it, we're currently re-considering continuing this (AURORA) series, since you were also able to run it on your phone, without loosing quality.
We hope you don't get any future issues with the model, and if you do, please tell us, and we'll do our best to fix it :)
Have a great day!
Of all the TinyLlama models that I downloaded to date (September 1, 2024), I could say that it was the best model that I was lucky enough to download for my cell phone, even surpassing the QWEN 2 in several aspects.
The testing has been carried out in Spanish, in which the level of responses is surprising and the great work and effort put in by the development team in this work is evident, even in the presentation of the model where it is precisely detailed what data to put in for its correct operation, which almost no one does.
It's a shame that its development has stopped.
Thanks.
Greetings once again, @Novaciano
Following internal restructuring, we are rebranding our AURORA series under the new name 'Intellect'. Beta versions are currently being rolled out, the next major V0.3 version expected to be open-sourced day after tomorrow, pending completion of ongoing evaluations The new series has a increase in size (1.1B => 1.6B) and is being trained on a substantial amount of high-quality data, both open-source and proprietary.
If you are interested, you may download the models from our profile. We welcome any feedback.
Wishing you a wonderful day!
Ah… I am pleased to know that the project has evolved, just as I wished it to. The renaming of AURORA to Intellect is an interesting move for an increase in parameter size, though the true value lies beyond mere rebranding… as you once managed during the times of AURORA. A larger scale and improved data quality… a necessary progress that I am eager to observe, especially considering the arrival of Llama 3.2 and the GRPO training methods.
I shall watch with interest the fruits of this new iteration. If you have achieved something truly noteworthy, I will discover it and let you know. Until then, keep striving—true intellect is not solely defined by the accumulation of data… but by the transcendence of what is simple and efficient.
Thank you for reaching out to me once again.
My best regards from afar.
Sincerely,
@Novaciano