mradermacher

/

model_requests

like 171

NotImplementedError: BPE pre-tokenizer was not recognized - update get_vocab_base_pre()

failed yesterday Sun May 24 15:05:34 2026

Do you have any idea why @Naphula ?

The model safetensors are almost uploaded, only 3 to go

(it fails often when set up to automated upload)

This should be ready to quantize now @RichardErkhov .

It's queued!

You can check for progress at http://hf.tst.eu/status.html or regularly check the model
summary page at https://hf.tst.eu/model#Morax-24B-v2-GGUF for quants to appear.

This model is unlucky cursed!

error/1 bpe-pt missing (77598ca9…)

This model is unlucky cursed!

error/1 bpe-pt missing (77598ca9…)

Any luck fixing it? I only tested Q8_0

I assume tokenizer is broken, so probably it needs fixing

Not sure what went wrong or how to fix it (not having issues on my end), but if it helps I can upload the F16 GGUF to requantize from

It requires some manual stuff which I dont exactly remember, but if it works with main llama cpp then we can try

In case there are further troubles I'm quantizing now the Q3KM, IQ4_XS, Q4KM, Q5KM, and Q6K. Imatrix can be generated later also if needed

If any special size is needed let me know (IQ4NL , Q5_1 are used rarely)

For 24B models I use imatrix Q4_0 for NPU.

Regular Q4_0 is queued. imatrix might take a few days but i'll try to get to those soon too

Thank you!