Update README.md
Browse files
README.md
CHANGED
@@ -17,6 +17,7 @@ language:
|
|
17 |
---
|
18 |
|
19 |
llama.cpp conversion of https://huggingface.co/euclaise/Memphis-scribe-3B
|
|
|
20 |
|
21 |
|
22 |
### ppl (512 wiki.test)
|
@@ -25,4 +26,6 @@ llama.cpp conversion of https://huggingface.co/euclaise/Memphis-scribe-3B
|
|
25 |
| f16(baseline) | 9.9437 +/- 0.07019 |
|
26 |
| q8_0 | 9.9474 +/- 0.07022 |
|
27 |
| q5_k_m | 10.0347 +/- 0.07091 |
|
28 |
-
| q4_k_m | 10.1192 +/- 0.07152 |
|
|
|
|
|
|
17 |
---
|
18 |
|
19 |
llama.cpp conversion of https://huggingface.co/euclaise/Memphis-scribe-3B
|
20 |
+
except for f16 and q8_0, every quant is using the imatrix from wiki-train
|
21 |
|
22 |
|
23 |
### ppl (512 wiki.test)
|
|
|
26 |
| f16(baseline) | 9.9437 +/- 0.07019 |
|
27 |
| q8_0 | 9.9474 +/- 0.07022 |
|
28 |
| q5_k_m | 10.0347 +/- 0.07091 |
|
29 |
+
| q4_k_m | 10.1192 +/- 0.07152 |
|
30 |
+
| iq2_xs | 16.6174 +/- 0.11807 |
|
31 |
+
| iq2_xxs | 22.6462 +/- 0.16226 |
|