metalglot

2 maanden geledendiscuss

Open source repo: https://github.com/grctest/finetuned-gemmatranslate-cy

5% of the fine-tuning took 40 minutes and cost a couple dollars to prove the process works.

Looking forwards to Flash Attention v4 to leave beta, to test fine-tuning performance on a B200 on the cloud, probably a few months away it seems?

What languages would you train TranslateGemma to be able to translate? I was originally thinking about klingon but the available datasets seemed a bit lacking..

metalglot

3 maanden geledendiscuss

Bets placed on misinformation spread on a friday that's outdated by saturday morning. Turning into a weekend tradition at this point..

metalglot

3 maanden geledendiscuss

Not if you use local models, only if you're using cloud llm's

metalglot

3 maanden geledendiscuss

Any moonbase will seriously run into this problem. Nightmare!

metalglot

3 maanden geledendiscuss

Check it out! It simplifies the setup of docker images, containers, downloading models and running multiple translation tasks!

Are there any other translation tasks you think have been missed out and should be supported?

If you're a builder then check out the opensource backend on GitHub: https://github.com/grctest/fastapi-gemma-translate

Submissions

Fine-Tuning TranslateGemma-4B for Better Welsh Translations on an H200 GPU

Comparing the best open source TranslateGemma projects

[untitled]

comments