Views
No views yet
./llama-cli -m Qalb-1.0-8B-Instruct-Q4_0.gguf -ngl 35 -c 8192 --color -cnv -p "سلام! کیسے مدد کر سکتا ہوں؟"| Setting | Value | Comment |
|---|---|---|
| Quantization | Q4_0 | Fast, ~4.5 GB VRAM / RAM |
| GPU layers (-ngl) | 30–50 | Adjust according to your GPU VRAM |
| Context size (-c) | 4096–8192 | Depends on use-case |
| Temperature | 0.7–0.9 | Higher = more creative |
| Repeat penalty | 1.1–1.15 | Helps reduce repetition |