File size: 1,460 Bytes
40c87cb 95d169b 41bfe1f d5e5b2a 40c87cb 41bfe1f 40c87cb e26df7a 41bfe1f 40c87cb 4b32e31 41bfe1f 40c87cb 41bfe1f 40c87cb 9538930 8a5f65d 9538930 7825a10 d5e5b2a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 |
---
library_name: transformers
tags:
- trl
- sft
base_model:
- meta-llama/Llama-3.2-1B-Instruct
datasets:
- ngxson/MiniThinky-dataset
new_version: ngxson/MiniThinky-v2-1B-Llama-3.2
---
# MiniThinky 1B
>[!IMPORTANT]
> **There is a newer checkpoint for this model, [click here](https://huggingface.co./ngxson/MiniThinky-v2-1B-Llama-3.2)**
My first trial to fine tune a small model to add reasoning capability.
Link to GGUF version: [click here](https://huggingface.co./ngxson/MiniThinky-1B-Llama-3.2-Q8_0-GGUF)
Chat template is the same with llama 3, but the response will be as follow:
```
<|thinking|>{thinking_process}
<|answer|>
{real_answer}
```
## IMPORTANT: System message
The model is **very sensitive** to system message. Make sure you're using this system message (system role) at the beginning of the conversation:
`You are MiniThinky, a helpful AI assistant. You always think before giving the answer. Use <|thinking|> before thinking and <|answer|> before giving the answer.`
## Q&A
**Hardware used to trained it?**
I used a HF space with 4xL40S, trained for 5 hours. Eval loss is about 0.8
**Benchmark?**
I don't have time to do it alone. If you can help, please open a discussion!
**Can it count number of "r" in "raspberry"?**
Unfortunately no
**Other things that I can tune?**
Maybe lower temperature, or set top_k=1
---
TODO: include more info here + maybe do some benchmarks? (Plz add a discussion if you're interested) |