Please make a heavily quantized version like you did with R1
#1
by
Blazgo
- opened
Do the same thing you did to make that 1.58 bit R1:
https://unsloth.ai/blog/deepseekr1-dynamic
Or if it doesn't need big GPUs tell me how to do it.
Blazgo
changed discussion status to
closed
Do the same thing you did to make that 1.58 bit R1:
https://unsloth.ai/blog/deepseekr1-dynamic
Or if it doesn't need big GPUs tell me how to do it.
We did it's here: https://huggingface.co/unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF