Kim JaeCheol
commited on
Commit
•
322906a
1
Parent(s):
0c3c0d9
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- ko
|
5 |
+
pipeline_tag: text-generation
|
6 |
+
---
|
7 |
+
## Prompt Tempalte
|
8 |
+
It follows Alpaca format.
|
9 |
+
```
|
10 |
+
### 질문: {instruction}
|
11 |
+
### 답변: {output}
|
12 |
+
```
|
13 |
+
### Implementation Code
|
14 |
+
```
|
15 |
+
import troch
|
16 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
17 |
+
|
18 |
+
model = AutoModelForCausalLM.fron_pretrained("Ja3ck/Mistral-instruct-IPO-Y24-v1", return_dict=True, torch_dtype=torch.bfloat16, device_map="auto")
|
19 |
+
tokenizer = AutoTokenizer.from_pretrained("Ja3ck/Mistral-instruct-IPO-Y24-v1", use_fast=True)
|
20 |
+
tokenizer.pad_token = tokenizer.unk_token
|
21 |
+
tokenizer.pad_token_id = tokenizer.unk_token_id
|
22 |
+
tokenizer.padding_side = "left"
|
23 |
+
|
24 |
+
def gen(x):
|
25 |
+
x_ = f"### 질문: {x.strip()} ### 답변: "
|
26 |
+
inputs = tokenizer(x_, return_tensor='pt')
|
27 |
+
input_ids = inputs['input_ids'].cuda()
|
28 |
+
generation_output = model.generate(
|
29 |
+
pad_token_id = tokenizer.pad_token_id,
|
30 |
+
temperature=0.1,
|
31 |
+
top_p=1,
|
32 |
+
top_k=50,
|
33 |
+
num_beams=1,
|
34 |
+
repetition_penalty=1.13,
|
35 |
+
do_sample=True,
|
36 |
+
),
|
37 |
+
return_dict_in_generate=True,
|
38 |
+
output_scores=True,
|
39 |
+
max_new_tokens=1024
|
40 |
+
)
|
41 |
+
for seq in generation_output.sequences:
|
42 |
+
output = tokenizer.decode(seq)
|
43 |
+
print(output.split("### 답변: ")[1].strip())
|
44 |
+
|
45 |
+
gen("안녕하세요?")
|
46 |
+
```
|