3 ms·
I suspect your "original gemma3:27b" was a quantized model since the non-quantized (16bit) version needs around 54gb.
by nolist_policy 2y ago
I suspect your "original gemma3:27b" was a quantized model since the non-quantized (16bit) version needs around 54gb.