Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,62 @@
|
|
1 |
-
---
|
2 |
-
license: gpl-3.0
|
3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: gpl-3.0
|
3 |
+
---
|
4 |
+
# NanoLM-70M-Instruct-v1
|
5 |
+
|
6 |
+
|
7 |
+
English | [简体中文](README_zh-CN.md)
|
8 |
+
|
9 |
+
|
10 |
+
## Introduction
|
11 |
+
|
12 |
+
In order to explore the potential of small models, I have attempted to build a series of them, which are available in the [NanoLM Collections](https://huggingface.co/collections/Mxode/nanolm-66d6d75b4a69536bca2705b2).
|
13 |
+
|
14 |
+
This is NanoLM-70M-Instruct-v1. The model currently supports **English only**.
|
15 |
+
|
16 |
+
|
17 |
+
|
18 |
+
## Model Details
|
19 |
+
|
20 |
+
The tokenizer and model architecture of NanoLM-70M-Instruct-v1 are the same as [SmolLM-135M](https://huggingface.co/HuggingFaceTB/SmolLM-135M), but the number of layers has been reduced from 30 to 12.
|
21 |
+
|
22 |
+
As a result, NanoLM-70M-Instruct-v1 has only 70 million parameters.
|
23 |
+
|
24 |
+
Despite this, NanoLM-70M-Instruct-v1 still demonstrates instruction-following capabilities.
|
25 |
+
|
26 |
+
|
27 |
+
|
28 |
+
## How to use
|
29 |
+
|
30 |
+
```python
|
31 |
+
import torch
|
32 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
33 |
+
|
34 |
+
model_path = 'Mxode/NanoLM-70M-Instruct-v1'
|
35 |
+
|
36 |
+
model = AutoModelForCausalLM.from_pretrained(model_path).to('cuda:0', torch.bfloat16)
|
37 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path)
|
38 |
+
|
39 |
+
|
40 |
+
text = "Why is it important for entrepreneurs to prioritize financial management?"
|
41 |
+
prompt = tokenizer.apply_chat_template(
|
42 |
+
[
|
43 |
+
{'role': 'system', 'content': 'You are a helpful assistant.'},
|
44 |
+
{'role': 'user', 'content': text}
|
45 |
+
],
|
46 |
+
add_generation_prompt=True,
|
47 |
+
tokenize=True,
|
48 |
+
return_tensors='pt'
|
49 |
+
).to('cuda:0')
|
50 |
+
|
51 |
+
|
52 |
+
outputs = model.generate(
|
53 |
+
prompt,
|
54 |
+
max_new_tokens=1024,
|
55 |
+
do_sample=True,
|
56 |
+
temperature=0.7,
|
57 |
+
repetition_penalty=1.1,
|
58 |
+
eos_token_id=tokenizer.eos_token_id,
|
59 |
+
)
|
60 |
+
response = tokenizer.decode(outputs[0])
|
61 |
+
print(response)
|
62 |
+
```
|