Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,23 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
This is [Allen AI's open-instruct dataset.](https://github.com/allenai/open-instruct)
|
| 6 |
+
|
| 7 |
+
It is used to train the Tulu family of models.
|
| 8 |
+
|
| 9 |
+
https://huggingface.co/allenai/tulu-7b
|
| 10 |
+
https://huggingface.co/allenai/tulu-13b
|
| 11 |
+
https://huggingface.co/allenai/tulu-30b
|
| 12 |
+
https://huggingface.co/allenai/tulu-65b
|
| 13 |
+
|
| 14 |
+
I have done the following:
|
| 15 |
+
|
| 16 |
+
1) Download the open-instruct repo
|
| 17 |
+
2) Execute the scripts/prepare_train_data.sh modified to download the "unfiltered" version of sharegpt dataset
|
| 18 |
+
3) Merged data/processed/**/*.jsonl into a single "open-instruct.jsonl"
|
| 19 |
+
4) Executed my "remove_refusals.py" against that "open-instruct.jsonl" to produce a "open-instruct-uncensored.jsonl"
|
| 20 |
+
|
| 21 |
+
I am currently training this "open-instruct-uncensored.jsonl" to a new model series named ehartford/tulu-uncensored
|
| 22 |
+
|
| 23 |
+
More info to come.
|