Update README.md
Browse files
README.md
CHANGED
|
@@ -19,3 +19,52 @@ configs:
|
|
| 19 |
- split: train
|
| 20 |
path: data/train-*
|
| 21 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
- split: train
|
| 20 |
path: data/train-*
|
| 21 |
---
|
| 22 |
+
# Dolly 15k Curated
|
| 23 |
+
|
| 24 |
+
## Dataset Details
|
| 25 |
+
|
| 26 |
+
### Dataset Description
|
| 27 |
+
|
| 28 |
+
A filtered and curated dataset version of https://huggingface.co/datasets/databricks/databricks-dolly-15k. Saved in HF Chat format. The result is a high quality dataset for SFT.
|
| 29 |
+
|
| 30 |
+
- **Created by:** [dctanner](https://huggingface.co/dctanner) and the team at [Sablo AI](https://sablo.ai)
|
| 31 |
+
- **License:** CC BY-SA 3.0
|
| 32 |
+
|
| 33 |
+
## Dataset Structure
|
| 34 |
+
|
| 35 |
+
We structure the dataset using the format commonly used as input into [Hugging Face Chat Templates](https://huggingface.co/docs/transformers/chat_templating):
|
| 36 |
+
|
| 37 |
+
```
|
| 38 |
+
[
|
| 39 |
+
{"role": "user", "content": "Hello, how are you?"},
|
| 40 |
+
{"role": "assistant", "content": "I'm doing great. How can I help you today?"}
|
| 41 |
+
]
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
## Dataset Creation
|
| 45 |
+
|
| 46 |
+
### Source Data
|
| 47 |
+
|
| 48 |
+
- **Source Dataset:** https://huggingface.co/datasets/databricks/databricks-dolly-15k
|
| 49 |
+
|
| 50 |
+
#### Data Collection and Processing
|
| 51 |
+
|
| 52 |
+
We started with https://huggingface.co/datasets/databricks/databricks-dolly-15k.
|
| 53 |
+
|
| 54 |
+
Filtering and curation was done to remove conversations with:
|
| 55 |
+
- Duplicate or very similar responses
|
| 56 |
+
- Responses where the AI was actually responding like a person (present in this dataset as the responses are created by humans pretending to be an AI, and no everyone followed these instructions closely)
|
| 57 |
+
- Profanity or inappropriate responses for an AI
|
| 58 |
+
- Very short response lengths (often below 50 or 200 characters)
|
| 59 |
+
|
| 60 |
+
# License
|
| 61 |
+
|
| 62 |
+
- **License:** CC BY-SA 3.0
|
| 63 |
+
|
| 64 |
+
This dataset is usable for commercial purposes. Certain categories of material in the dataset include materials from the following sources, licensed under the CC BY-SA 3.0 license:
|
| 65 |
+
- Wikipedia (various pages) - https://www.wikipedia.org/ - Copyright © Wikipedia editors and contributors.
|
| 66 |
+
- Databricks (https://www.databricks.com) - Copyright © Databricks
|
| 67 |
+
|
| 68 |
+
# Contact
|
| 69 |
+
|
| 70 |
+
Created by [dctanner](https://huggingface.co/dctanner) and the team at [Sablo AI](https://sablo.ai)
|