prithivMLmods

/

Sombrero-QwQ-32B-Elite10-Broken

like 13

This model is tuned based on streamlined memory optimization, which reduces the object state of inference in less time by avoiding unnecessary schema. It is specifically designed for Math, though it is an experimental attempt with different settings that resulted in these three models. Let's see how it performs in the coming days. It works best in [English and Chinese].

@mmbela

From the description Elite10 looks modified for different purposes than the other two: text process, summarization. I asked this because of this and I thought there are changes in the Multilingual Proficiency too. I knew that these qwen models are for two languages mainly. Thanks.

@mmbela
Yes, it does support other languages mentioned, but it works best in English and Chinese. It's also specifically focused on improving mathematical problem-solving and enhancing contextual comprehension.

You can find 2-bit Q2_K, 4-bit Q4 K_S, and Q4 K_M here:
@mmbela - in a while

https://huggingface.co/prithivMLmods/Sombrero-QwQ-32B-Elite10-GGUF

Really thanks, I'll try it. Was it my fault? I got this message at the end of the conversion by llama.cpp and I was not able to load the gguf:
INFO:hf-to-gguf:Model successfully exported to /Users/bela/Desktop/model/qwen/somb-elite10/smobero-qwq-elite10-f16.gguf

I downloaded the Q4K_M. The error message in LM Studio:

🥲 Failed to load the model

Failed to load model

error loading model: check_tensor_dims: tensor 'token_embd.weight' has wrong shape; expected  5120, 152064, got  5120, 151665,     1,     1

I have no problem with Eliti11 gguf.

16bit

https://huggingface.co/DevQuasar/prithivMLmods.Sombrero-QwQ-32B-Elite10-GGUF

I'm facing the same error. I'm fixing it right now!
I'll upload the new checkpoints soon.

Thank you for raising the issue.
@mmbela