3 ms·
Maybe model is sensitive to quantization, by default ollama quantize it significantly.
by pzo 2y ago
Maybe model is sensitive to quantization, by default ollama quantize it significantly.
- tarruda 2y agoI tried ollama fp16 and it had the same issues.