search

RichardErkhov / quant_request

You can request more quants here
1 stars 0 forks source link

requesting quant of an existing model; #1

Open evenfog opened 5 months ago

evenfog commented 5 months ago

https://huggingface.co/Kooten/DaringMaid-20B-V1.1

^ That model has much better results than the original DaringMaid 20b from Kooten in terms of sticking with context. I have wanted to use a Q8_0 version for high fidelity use and only can find lesser quants in GGUF for that model, but it's probably the best 20b that I can find in terms of creative writing and adhering to complex situational contexts, want to put it to use at the best possible level in GGUF format!

RichardErkhov commented 5 months ago

Will be done soon

evenfog commented 5 months ago

Nice thanks king!

RichardErkhov commented 5 months ago

https://huggingface.co/RichardErkhov/Kooten_-_DaringMaid-20B-V1.1-gguf

on the way =) will take some time because running multiple models in parallel and my internet does not support that decision

RichardErkhov commented 5 months ago

anything else you want me to quant?

razvanab commented 4 months ago

I notice that you are making dolphin variants of models too on HuggingFace. Do you think you can also make a dolphin version of the Microsoft Phi-3 mini? https://huggingface.co/microsoft/Phi-3-mini-4k-instruct

RichardErkhov commented 4 months ago

I notice that you are making dolphin variants of models too on HuggingFace. Do you think you can also make a dolphin version of the Microsoft Phi-3 mini? https://huggingface.co/microsoft/Phi-3-mini-4k-instruct

I dont finetune at all. I mean if you provide me with step-by-step guide on finetuning, I can try to do that for you. I do only quantization for now, but will be interesting to try finetuning

razvanab commented 4 months ago

Ok, I see. I am trying to learn to fine-tune LLM too. Huggingface Autotune and LLaMA-Factory are what people recommend. Thanks.