Commit
•
b8d6a13
1
Parent(s):
3d513f8
Update README.md
Browse files
README.md
CHANGED
|
@@ -8,7 +8,15 @@ tags:
|
|
| 8 |
|
| 9 |
---
|
| 10 |
A zero training self-merge test of [Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R](https://huggingface.co/Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R) using settings mentioned on [mistral-11b-slimorca](https://huggingface.co/chargoddard/mistral-11b-slimorca)<br>
|
|
|
|
| 11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
<b><u>Tools used / version</u></b><br>
|
| 14 |
<b>Mergekit:</b> <i>c93c9bb</i><br>
|
|
|
|
| 8 |
|
| 9 |
---
|
| 10 |
A zero training self-merge test of [Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R](https://huggingface.co/Salesforce/SFR-Iterative-DPO-LLaMA-3-8B-R) using settings mentioned on [mistral-11b-slimorca](https://huggingface.co/chargoddard/mistral-11b-slimorca)<br>
|
| 11 |
+
It's....not dumber I guess 🤷♀️
|
| 12 |
|
| 13 |
+
Simple PPL comparison<br>
|
| 14 |
+
<code>
|
| 15 |
+
<i>perplexity.exe -[MODEL] -f wiki.test.raw -b 512 -ngl 99</i><br>
|
| 16 |
+
<i>SFR-Iterative-DPO-LLaMA-3-8B-R-F16.gguf</i> - Final estimate: <b>PPL = 7.0279 +/- 0.04493</b><br>
|
| 17 |
+
<i>SFR-Iterative-DPO-LLaMA-3-11.5B-R-Q6_K.gguf</i> - Final estimate: <b>PPL = 7.0500 +/- 0.04516</b><br>
|
| 18 |
+
<i>Meta-Llama-3-8B-Instruct-Q6_K</i> - Final estimate: <b>PPL = 8.4727 +/- 0.06308</b>
|
| 19 |
+
</code>
|
| 20 |
|
| 21 |
<b><u>Tools used / version</u></b><br>
|
| 22 |
<b>Mergekit:</b> <i>c93c9bb</i><br>
|