Plz quants for Firefly

#412
by invisietch - opened

Please could you add https://huggingface.co/invisietch/MiS-Firefly-v0.1-22B to your queue?

It may need tokenizer.model from Mistral Small repo (it did for me locally).

Great, I hate hacks like that :(

It's added, and I will replace the file when needed.

Cheers!

mradermacher changed discussion status to closed

I assume you mean https://huggingface.co/mistralai/Mistral-Small-Instruct-2409 - I don't have access to that repo at the moment, any alternative?

Do you have a Discord? I can send the file. Mine is the same as username here.

gguf-my-repo seems to quant it without the file but my local llama.cpp wouldn't.

Ah, ok, I'll try without any hacks first.

If you give the model a try, let me know what you think. I get 70B vibes from it especially in creative writing.

Worked fine, should be done... very soon.

Already added to model card!

Too fast :) But seriously, you should link the imatrix quants instead once finished.

Too fast :) But seriously, you should link the imatrix quants instead once finished.

I linked both ;)

Hi, apologies. The model worked well at Q8 but there was a bug with quantized models (not your fault, happened when I tried making Q6 locally too).

It's fixed in https://huggingface.co/invisietch/MiS-Firefly-v0.2-22B -- would you mind doing GGUF for this when its uploaded in a couple of hours?

Sure, just notify me when it's fully uploaded :)

It's up, and thanks!

Aaand its queued :)

Sign up or log in to comment