teilomillet
commited on
Commit
•
d13cdbc
1
Parent(s):
8bb5a2d
Update README.md
Browse files
README.md
CHANGED
@@ -9,4 +9,28 @@ widget:
|
|
9 |
- text: "<s> [|User|] Comment faire un bon plat ? </s>[|Assistant|]"
|
10 |
---
|
11 |
SFT on a synthetic custom (french) dataset (2k), from general question answering, problem solving to code question.
|
12 |
-
It's a POC.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
- text: "<s> [|User|] Comment faire un bon plat ? </s>[|Assistant|]"
|
10 |
---
|
11 |
SFT on a synthetic custom (french) dataset (2k), from general question answering, problem solving to code question.
|
12 |
+
It's a POC.
|
13 |
+
|
14 |
+
|
15 |
+
```python
|
16 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
17 |
+
from peft import PeftModel
|
18 |
+
import torch
|
19 |
+
|
20 |
+
model = AutoModelForCausalLM.from_pretrained(
|
21 |
+
"teilomillet/MiniMerlin-3B",
|
22 |
+
revision="0.1",
|
23 |
+
return_dict=True,
|
24 |
+
torch_dtype=torch.bfloat16,
|
25 |
+
device_map='auto'
|
26 |
+
)
|
27 |
+
|
28 |
+
tokenizer = AutoTokenizer.from_pretrained("teilomillet/MiniMerlin-3B")
|
29 |
+
tokenizer.pad_token = tokenizer.eos_token
|
30 |
+
|
31 |
+
text = "[|User|] Comment faire un bon plat ? </s>[|Assistant|]"
|
32 |
+
inputs = tokenizer(text, return_tensors="pt").to(0)
|
33 |
+
|
34 |
+
outputs = model.generate(**inputs, max_new_tokens=800)
|
35 |
+
print(tokenizer.decode(outputs[0], skip_special_tokens=False))
|
36 |
+
```
|