Instructions to use TheBloke/mpt-30B-GGML with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use TheBloke/mpt-30B-GGML with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("TheBloke/mpt-30B-GGML", dtype="auto") - Notebooks
- Google Colab
- Kaggle
GPTQ version for italians
Hi there @TheBloke , first thank you for what you are doing, you are contributing so much to democratize these models.
I'am here to ask you if you could release a GPTQ version of this model (MPT-30B) because i noticed that it particularly works well for languages that are not english (in my case Italian) and it works so much better than Llama-2 also for these use cases.
This could really help the opensource community of non-english people (and we are a lot) cause it really makes it possible to use and fine tune it.
If it's not possible I could also appreciate if you link me some repository where I can do it myself.
Thank you very much for everything you are doing.
I could be wrong but I think there was some discussion about this when mpt came out but the architecture was different so none of the existing processes worked?