Qwen2.5 Math 7B Instruct
Hi, can you please create quants for Qwen2.5 Math 7b Instruct? Bartowski already has quants; however, he uses an imatrix which could potentially harm the model as it is multilingual (English and Chinese).
Also, off topic question, are you related to https://x.com/ra_der_macher?
He already quantized all Qwen 2.5 models but keep in mind that some will have to be quantized due to them being originally uploaded with the wrong tokenizer. For more information, please look at: https://huggingface.co/mradermacher/model_requests/discussions/297. Qwen2.5-Math-7B is currently likely affected by this issue.
Oh no you are right we somehow forget Qwen/Qwen2.5-Math-7B-Instruct. No idea how this happened as I was sure I listed all Qwen 2.5 models. I must have somehow missed it.
@isr431 what nico said, and, no, mradermacher only has this account on huggingface, and it's just a pseudonym, so any other existing mradermacher is somebody else.
i still have to go through the list and redo models, but likely i will redo all models <72b, and a few affected 72b models.
Anyway, I have queued https://huggingface.co/Qwen/Qwen2.5-Math-7B-Instruct because I really don't have that. Which bis strange, because I actually used a glob pattern to generate the urls to submit, I thought. How can one be missing.
Ah, yeah because I only did it for the non-math-models:llmjob a s i 1500 https://huggingface.co/Qwen/Qwen2.5-{0.5,1.5,3,7,14,32,72}B{,-Instruct}
@nicoboss I'd say qwen2.5-32b might be an excellent candidate as a medium-size benchmark model for all quants.