zR
commited on
Commit
•
a1817f2
1
Parent(s):
6b3e945
check
Browse files- README.md +60 -0
- README_zh.md +44 -0
- generation_config.json +0 -10
README.md
ADDED
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
frameworks:
|
3 |
+
- Pytorch
|
4 |
+
license: other
|
5 |
+
license_name: glm-4
|
6 |
+
license_link: LICENSE
|
7 |
+
pipeline_tag: text-generation
|
8 |
+
tags:
|
9 |
+
- glm
|
10 |
+
- edge
|
11 |
+
inference: false
|
12 |
+
---
|
13 |
+
|
14 |
+
# GLM-Edge-4B-Chat
|
15 |
+
|
16 |
+
中文阅读, 点击[这里](README_zh.md)
|
17 |
+
|
18 |
+
## Inference with Transformers
|
19 |
+
|
20 |
+
### Installation
|
21 |
+
|
22 |
+
Install the transformers library from the source code:
|
23 |
+
|
24 |
+
```shell
|
25 |
+
pip install git+https://github.com/huggingface/transformers.git
|
26 |
+
```
|
27 |
+
|
28 |
+
### Inference
|
29 |
+
|
30 |
+
```python
|
31 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
32 |
+
|
33 |
+
MODEL_PATH = "THUDM/glm-edge-4b-chat"
|
34 |
+
|
35 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
|
36 |
+
model = AutoModelForCausalLM.from_pretrained(MODEL_PATH, device_map="auto")
|
37 |
+
|
38 |
+
message = [{"role": "user", "content": "hello!"}]
|
39 |
+
|
40 |
+
inputs = tokenizer.apply_chat_template(
|
41 |
+
message,
|
42 |
+
return_tensors="pt",
|
43 |
+
add_generation_prompt=True,
|
44 |
+
return_dict=True,
|
45 |
+
).to(model.device)
|
46 |
+
|
47 |
+
generate_kwargs = {
|
48 |
+
"input_ids": inputs["input_ids"],
|
49 |
+
"attention_mask": inputs["attention_mask"],
|
50 |
+
"max_new_tokens": 128,
|
51 |
+
"do_sample": False,
|
52 |
+
}
|
53 |
+
out = model.generate(**generate_kwargs)
|
54 |
+
print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
|
55 |
+
|
56 |
+
```
|
57 |
+
|
58 |
+
## License
|
59 |
+
|
60 |
+
The usage of this model’s weights is subject to the terms outlined in the [LICENSE](LICENSE).
|
README_zh.md
ADDED
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# GLM-Edge-4B-Chat
|
2 |
+
|
3 |
+
## 使用 transformers 库进行推理
|
4 |
+
|
5 |
+
### 安装
|
6 |
+
|
7 |
+
请安装源代码的transformers库。
|
8 |
+
|
9 |
+
```shell
|
10 |
+
pip install git+https://github.com/huggingface/transformers.git
|
11 |
+
```
|
12 |
+
### 推理
|
13 |
+
|
14 |
+
```python
|
15 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
16 |
+
|
17 |
+
MODEL_PATH = "THUDM/glm-edge-4b-chat"
|
18 |
+
|
19 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
|
20 |
+
model = AutoModelForCausalLM.from_pretrained(MODEL_PATH, device_map="auto")
|
21 |
+
|
22 |
+
message = [{"role": "user", "content": "hello!"}]
|
23 |
+
|
24 |
+
inputs = tokenizer.apply_chat_template(
|
25 |
+
message,
|
26 |
+
return_tensors="pt",
|
27 |
+
add_generation_prompt=True,
|
28 |
+
return_dict=True,
|
29 |
+
).to(model.device)
|
30 |
+
|
31 |
+
generate_kwargs = {
|
32 |
+
"input_ids": inputs["input_ids"],
|
33 |
+
"attention_mask": inputs["attention_mask"],
|
34 |
+
"max_new_tokens": 128,
|
35 |
+
"do_sample": False,
|
36 |
+
}
|
37 |
+
out = model.generate(**generate_kwargs)
|
38 |
+
print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
|
39 |
+
|
40 |
+
```
|
41 |
+
|
42 |
+
## 协议
|
43 |
+
|
44 |
+
本模型的权重的使用则需要遵循 [LICENSE](LICENSE)。
|
generation_config.json
DELETED
@@ -1,10 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"_from_model_config": true,
|
3 |
-
"eos_token_id": [
|
4 |
-
59246,
|
5 |
-
59253,
|
6 |
-
59255
|
7 |
-
],
|
8 |
-
"pad_token_id": 59246,
|
9 |
-
"transformers_version": "4.47.0.dev0"
|
10 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|