mradermacher
commited on
Update README.md
Browse files
README.md
CHANGED
@@ -40,7 +40,7 @@ The only fix seems to be to delete the repo, which unfortunately also deletes th
|
|
40 |
|
41 |
The quant types I currently do regularly are:
|
42 |
|
43 |
-
- static: (F16) Q8_0
|
44 |
- imatrix: Q2_K Q4_K_S IQ3_XXS Q3_K_M Q4_K_M IQ2_M Q6_K IQ4_XS Q3_K_S Q3_K_L Q5_K_S Q5_K_M Q4_0 IQ3_XS IQ3_S IQ3_M IQ2_XXS IQ2_XS IQ2_S IQ1_M IQ1_S (Q4_0_4_4 Q4_0_4_8 Q4_0_8_8)
|
45 |
|
46 |
And they are generally (but not always) generated in the order above, for which there are deep reasons.
|
@@ -54,6 +54,9 @@ this year. At least when multiple quant types are missing, as it is hard to just
|
|
54 |
for just one quant. If you want a quant form the above list and don't want to wait, feel free to request it and I will
|
55 |
prioritize it to the best of my abilities.
|
56 |
|
|
|
|
|
|
|
57 |
I specifically do not do Q2_K_S, because I generally think it is not worth it, and IQ4_NL, because it requires
|
58 |
a lot of computing and is generally completely superseded by IQ4_XS.
|
59 |
|
|
|
40 |
|
41 |
The quant types I currently do regularly are:
|
42 |
|
43 |
+
- static: (F16) Q8_0 Q4_K_S Q2_K Q6_K Q3_K_M Q3_K_S Q3_K_L Q4_K_M Q5_K_S Q5_K_M IQ4_XS
|
44 |
- imatrix: Q2_K Q4_K_S IQ3_XXS Q3_K_M Q4_K_M IQ2_M Q6_K IQ4_XS Q3_K_S Q3_K_L Q5_K_S Q5_K_M Q4_0 IQ3_XS IQ3_S IQ3_M IQ2_XXS IQ2_XS IQ2_S IQ1_M IQ1_S (Q4_0_4_4 Q4_0_4_8 Q4_0_8_8)
|
45 |
|
46 |
And they are generally (but not always) generated in the order above, for which there are deep reasons.
|
|
|
54 |
for just one quant. If you want a quant form the above list and don't want to wait, feel free to request it and I will
|
55 |
prioritize it to the best of my abilities.
|
56 |
|
57 |
+
The (static) IQ3 quants are no longer generated, as they consistently seem to result in *much* lower quality
|
58 |
+
quants than even static Q2_K, so it would be s disservice to offer them.
|
59 |
+
|
60 |
I specifically do not do Q2_K_S, because I generally think it is not worth it, and IQ4_NL, because it requires
|
61 |
a lot of computing and is generally completely superseded by IQ4_XS.
|
62 |
|