Update README.md
Browse files
README.md
CHANGED
@@ -16,11 +16,44 @@ Llama3-Chinese-8B-Instruct基于Llama3-8B中文微调对话模型,由Llama中
|
|
16 |
```
|
17 |
import transformers
|
18 |
import torch
|
|
|
|
|
19 |
model_id = "FlagAlpha/Llama3-Chinese-8B-Instruct"
|
20 |
|
21 |
pipeline = transformers.pipeline(
|
22 |
-
"text-generation",
|
|
|
|
|
|
|
23 |
)
|
24 |
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
```
|
|
|
16 |
```
|
17 |
import transformers
|
18 |
import torch
|
19 |
+
|
20 |
+
|
21 |
model_id = "FlagAlpha/Llama3-Chinese-8B-Instruct"
|
22 |
|
23 |
pipeline = transformers.pipeline(
|
24 |
+
"text-generation",
|
25 |
+
model=model_id,
|
26 |
+
model_kwargs={"torch_dtype": torch.float16},
|
27 |
+
device="cuda",
|
28 |
)
|
29 |
|
30 |
+
|
31 |
+
messages = [{"role": "system", "content": ""}]
|
32 |
+
|
33 |
+
messages.append(
|
34 |
+
{"role": "user", "content": "介绍一下机器学习"}
|
35 |
+
)
|
36 |
+
|
37 |
+
prompt = pipeline.tokenizer.apply_chat_template(
|
38 |
+
messages,
|
39 |
+
tokenize=False,
|
40 |
+
add_generation_prompt=True
|
41 |
+
)
|
42 |
+
|
43 |
+
terminators = [
|
44 |
+
pipeline.tokenizer.eos_token_id,
|
45 |
+
pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
|
46 |
+
]
|
47 |
+
outputs = pipeline(
|
48 |
+
prompt,
|
49 |
+
max_new_tokens=512,
|
50 |
+
eos_token_id=terminators,
|
51 |
+
do_sample=True,
|
52 |
+
temperature=0.6,
|
53 |
+
top_p=0.9
|
54 |
+
)
|
55 |
+
|
56 |
+
content = outputs[0]["generated_text"][len(prompt):]
|
57 |
+
|
58 |
+
print(content)
|
59 |
```
|