MaLA-LM

/

lucky52-bloom-7b1-no-52

like 0

Follow

MaLA-LM 20

Hi, we don't claim advantages over the original model, but we conducted an empirical study on multilingual instruction tuning the original model. Check out our paper for our findings https://arxiv.org/abs/2404.04850

Thanks for making the gguf. I have a series of instruction-finetuned models https://huggingface.co/collections/MaLA-LM/lucky52-660e5fd24a2ced4b334d63d6 If you would like to quantize them, I am happy to make a new collection.

I uploaded the quantized version to Jan and got no response. But I suspect that you've made versions for different languages, so you've marked them as multilingual. I don't know if that's true. If that's the case, I'd like to find a way for each model to match a specific language.

I don't have much experience. My interest is dictated by attempts to find a suitable model for translating subtitles for popular science films. Such content requires knowledge of terminological nomenclature, as well as its adaptation to the context of the target language of translation. In addition, the cultural context, adaptation of idioms and literary translation are important. This task goes beyond direct or literal translation.
Thanks