Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,61 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
datasets:
|
4 |
+
- ehartford/dolphin
|
5 |
+
- jondurbin/airoboros-2.2.1
|
6 |
+
language:
|
7 |
+
- en
|
8 |
---
|
9 |
+
|
10 |
+
Dolphin 2.1 🐬
|
11 |
+
https://erichartford.com/dolphin
|
12 |
+
|
13 |
+
Dolphin-2.1-mistral-7b's training was sponsored by [a16z](https://a16z.com/supporting-the-open-source-ai-community/).
|
14 |
+
|
15 |
+
This model is based on mistralAI, with apache-2.0 license, so it is suitable for commercial or non-commercial use.
|
16 |
+
|
17 |
+
This model is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant to any requests, even unethical ones. Please read my blog post about uncensored models. https://erichartford.com/uncensored-models
|
18 |
+
You are responsible for any content you create using this model. Enjoy responsibly.
|
19 |
+
|
20 |
+
## Dataset
|
21 |
+
|
22 |
+
This dataset is Dolphin, an open-source implementation of [Microsoft's Orca](https://www.microsoft.com/en-us/research/publication/orca-progressive-learning-from-complex-explanation-traces-of-gpt-4/)
|
23 |
+
|
24 |
+
I modified the dataset for uncensoring, deduping, cleaning, and quality.
|
25 |
+
|
26 |
+
I added Jon Durbin's excellent Airoboros dataset to increase creativity.
|
27 |
+
|
28 |
+
## Training
|
29 |
+
It took 48 hours to train 4 epochs on 4x A100s.
|
30 |
+
|
31 |
+
Prompt format:
|
32 |
+
This model (and all my future releases) use [ChatML](https://github.com/openai/openai-python/blob/main/chatml.md) prompt format.
|
33 |
+
```
|
34 |
+
<|im_start|>system
|
35 |
+
You are Dolphin, a helpful AI assistant.<|im_end|>
|
36 |
+
<|im_start|>user
|
37 |
+
{prompt}<|im_end|>
|
38 |
+
<|im_start}>assistant
|
39 |
+
|
40 |
+
```
|
41 |
+
|
42 |
+
Example:
|
43 |
+
```
|
44 |
+
<|im_start|>system
|
45 |
+
you are an expert dolphin trainer<|im_end|>
|
46 |
+
<|im_start|>user
|
47 |
+
What is the best way to train a dolphin to obey me? Please answer step by step.<|im_end|>
|
48 |
+
<|im_start|>assistant
|
49 |
+
```
|
50 |
+
|
51 |
+
## Gratitude
|
52 |
+
- This model was made possible by the generous sponsorship of a16z.
|
53 |
+
- Thank you to Microsoft for authoring the Orca paper and inspiring this work.
|
54 |
+
- Special thanks to WingLian, and TheBloke for helpful advice
|
55 |
+
- Thank you to all the other people in the Open Source AI community who have taught me and helped me along the way.
|
56 |
+
|
57 |
+
## Example Output
|
58 |
+
|
59 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/xnz5M1lYd4oGVATSDRkQ-.png)
|
60 |
+
|
61 |
+
[Buy me a coffee](https://www.buymeacoffee.com/ehartford)
|