Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
```
|
2 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
3 |
+
import torch
|
4 |
+
|
5 |
+
model_id = "mzbac/gemma-2-9b-grammar-correction"
|
6 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
7 |
+
model = AutoModelForCausalLM.from_pretrained(
|
8 |
+
model_id,
|
9 |
+
torch_dtype=torch.bfloat16,
|
10 |
+
device_map="auto",
|
11 |
+
)
|
12 |
+
|
13 |
+
messages = [
|
14 |
+
{
|
15 |
+
"role": "user",
|
16 |
+
"content": "Please correct, polish, or translate the text delimited by triple backticks to standard English\nText=```neither 经理或员工 has been informed about the meeting```",
|
17 |
+
},
|
18 |
+
]
|
19 |
+
input_ids = tokenizer.apply_chat_template(
|
20 |
+
messages, add_generation_prompt=True, return_tensors="pt"
|
21 |
+
).to(model.device)
|
22 |
+
|
23 |
+
terminators = [tokenizer.eos_token_id, tokenizer.convert_tokens_to_ids("<|im_end|>")]
|
24 |
+
|
25 |
+
outputs = model.generate(
|
26 |
+
input_ids,
|
27 |
+
max_new_tokens=256,
|
28 |
+
eos_token_id=terminators,
|
29 |
+
do_sample=True,
|
30 |
+
temperature=0.1,
|
31 |
+
)
|
32 |
+
response = outputs[0]
|
33 |
+
print(tokenizer.decode(response))
|
34 |
+
|
35 |
+
# <bos><start_of_turn>user
|
36 |
+
# Please correct, polish, or translate the text delimited by triple backticks to standard English
|
37 |
+
# Text=```neither 经理或员工 has been informed about the meeting```<end_of_turn>
|
38 |
+
# <start_of_turn>model
|
39 |
+
# Output=Neither the manager nor the employees have been informed about the meeting.<end_of_turn>
|
40 |
+
# <eos>
|