Update README.md
Browse files
README.md
CHANGED
|
@@ -12,7 +12,21 @@ tags:
|
|
| 12 |
base_model: unsloth/llama-3-8b-bnb-4bit
|
| 13 |
---
|
| 14 |
|
| 15 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
|
| 17 |
- **Developed by:** cookinai
|
| 18 |
- **License:** apache-2.0
|
|
|
|
| 12 |
base_model: unsloth/llama-3-8b-bnb-4bit
|
| 13 |
---
|
| 14 |
|
| 15 |
+
# Llama 3 finetuned on my TRRR-CoT Dataset
|
| 16 |
+
**cookinai/TRRR-CoT**
|
| 17 |
+
|
| 18 |
+
- This was an attempt at synthetically generating a CoT dataset and then finetuning it on a model to see its reuslts.
|
| 19 |
+
- From what I notice, when using the correct prompt template the model almost always ues the TRRR format, but I am still awaiting benchmark tests to see if this can improve anything
|
| 20 |
+
- TRR stand for:
|
| 21 |
+
|
| 22 |
+
1. **Think**, about your response
|
| 23 |
+
2. **Respond**, how you normally would
|
| 24 |
+
3. **Reflect**, on your response
|
| 25 |
+
4. **Respond**, again but this time use all the information you have now
|
| 26 |
+
|
| 27 |
+
- The mode usually tries to follow this format, it may mix it up a little but usually it almost always reflects in someway. Especially if you tell it to think step by step
|
| 28 |
+
|
| 29 |
+
- Intrestingly enough, when finetuned on mistral 7b, I could not get the model CoT at all, with only one epoch llama 3 got it instantly
|
| 30 |
|
| 31 |
- **Developed by:** cookinai
|
| 32 |
- **License:** apache-2.0
|