Update README.md
Browse files
README.md
CHANGED
@@ -29,4 +29,20 @@ The model is **very sensitive** to system message. Make sure you're using this s
|
|
29 |
You are MiniThinky, a helpful AI assistant. You always think before giving the answer. Use <|thinking|> before thinking and <|answer|> before giving the answer.
|
30 |
```
|
31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
32 |
TODO: include more info here + maybe do some benchmarks? (Plz add a discussion if you're interested)
|
|
|
29 |
You are MiniThinky, a helpful AI assistant. You always think before giving the answer. Use <|thinking|> before thinking and <|answer|> before giving the answer.
|
30 |
```
|
31 |
|
32 |
+
## Q&A
|
33 |
+
|
34 |
+
**Hardware used to trained it?**
|
35 |
+
I used a HF space with 4xL40S, trained for 5 hours. Eval loss is about 0.8
|
36 |
+
|
37 |
+
**Benchmark?**
|
38 |
+
I don't have time to do it alone. If you can help, please open a discussion!
|
39 |
+
|
40 |
+
**Can it count number of "r" in "raspberry"?**
|
41 |
+
Unfortunately no
|
42 |
+
|
43 |
+
**Other things that I can tune?**
|
44 |
+
Maybe lower temperature, or set top_k=1
|
45 |
+
|
46 |
+
---
|
47 |
+
|
48 |
TODO: include more info here + maybe do some benchmarks? (Plz add a discussion if you're interested)
|