Update README.md
Browse files
README.md
CHANGED
|
@@ -56,6 +56,8 @@ Quants Q4_0/Q5_0 for portable, phone and other devices.
|
|
| 56 |
|
| 57 |
Q8 is a maxed quant only, as imatrix has no effect on this quant.
|
| 58 |
|
|
|
|
|
|
|
| 59 |
Note that IQ1s performance is low, whereas IQ2s are passable.
|
| 60 |
|
| 61 |
More information on quants is in the document below "Highest Quality Settings / Optimal Operation Guide / Parameters and Samplers".
|
|
|
|
| 56 |
|
| 57 |
Q8 is a maxed quant only, as imatrix has no effect on this quant.
|
| 58 |
|
| 59 |
+
F16 is full precision, with the full "power" of the model.
|
| 60 |
+
|
| 61 |
Note that IQ1s performance is low, whereas IQ2s are passable.
|
| 62 |
|
| 63 |
More information on quants is in the document below "Highest Quality Settings / Optimal Operation Guide / Parameters and Samplers".
|