3 ms·
As far as I can tell Yi-34B is natively 16 bit float, the 8 bit version is quantized. https://huggingface.co/01-ai/Yi-34B#quantization https://huggingface.co/01
by modeless 3y ago
As far as I can tell Yi-34B is natively 16 bit float, the 8 bit version is quantized. https://huggingface.co/01-ai/Yi-34B#quantization https://huggingface.co/01-ai/Yi-34B#quantization