Quantization of Deepseek R1
@mradermacher Would be a great contribution to the community. You were the only one offering Imatrix versions of the large Deepseek models.
No worries we are already working on it. Please check https://huggingface.co/mradermacher/DeepSeek-V3-GGUF/discussions/2 for the discussion about it. Keep an eye on http://hf.tst.eu/status.html and expect first imatrix quants of DeepSeek-R1 to be available in a few days with DeepSeek-R1-Zero following later.
You were the only one offering Imatrix versions of the large Deepseek models.
Not that surprising considering how challenging computing the imatrix of such a massive model is. To compute the imatrix for such massive models we must connect multiple servers over RPC using 10 Gbit networking as my largest server only has 512 GiB of RAM. But in the end the effort is worth it as they are currently one of the best if not the best openly available large language models.