Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
pipeline_tag: text-generation
|
6 |
+
---
|
7 |
+
![palmer](https://huggingface.co/appvoid/palmer-001/resolve/main/palmer.jpeg)
|
8 |
+
# palmer
|
9 |
+
### a better base model
|
10 |
+
palmer is a series of ~1b parameters language models fine-tuned to be used as base models instead of using custom prompts for tasks. This means that it can be further fine-tuned on more data with custom prompts as usual or be used for downstream tasks as any base model you can get. The model has the best of both worlds: some "bias" to act as an assistant, but also the abillity to predict the next-word from its internet knowledge base. It's a 1.1b llama 2 model so you can use it with your favorite tools/frameworks.
|
11 |
+
`palmer-002` is a 1.3b parameters model.
|
12 |
+
|
13 |
+
### evaluation
|
14 |
+
|Model| ARC_C| HellaSwag| PIQA| Winogrande|
|
15 |
+
|------|-----|-----------|------|-------------|
|
16 |
+
|tinyllama-2t| 0.2807| 0.5463| 0.7067| 0.5683|
|
17 |
+
|palmer-001 | 0.2807| 0.5524| 0.7106| **0.5896**|
|
18 |
+
|sheared-1.3b| 0.2910| 0.5935| **0.7339**| 0.5809|
|
19 |
+
|palmer-002 | **0.3157**| **0.6022**| 0.7334| 0.5864|
|
20 |
+
|
21 |
+
### training
|
22 |
+
Training took ~5 P100 gpu hours. It was trained on 15,000 gpt-4 shuffled samples. palmer was fine-tuned using lower learning rates ensuring it keeps as much general knowledge as possible.
|
23 |
+
|
24 |
+
### prompt
|
25 |
+
```
|
26 |
+
no prompt
|
27 |
+
```
|
28 |
+
|
29 |
+
As you can notice, the model actually completes by default questions that are the most-likely to be asked, which is good because most people will use it to answer as a chatbot.
|
30 |
+
<a href="https://ko-fi.com/appvoid" target="_blank"><img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me A Coffee" style="height: 48px !important;width: 180px !important; filter: invert(70%);" ></a>
|