Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- zh
|
| 5 |
+
- en
|
| 6 |
+
pipeline_tag: text-generation
|
| 7 |
+
library_name: transformers
|
| 8 |
+
---
|
| 9 |
+
<div align="center">
|
| 10 |
+
<img src="https://github.com/OpenBMB/MiniCPM/blob/main/assets/minicpm_logo.png?raw=true" width="500em" ></img>
|
| 11 |
+
</div>
|
| 12 |
+
|
| 13 |
+
## Usage
|
| 14 |
+
### Prebuilt [AutoAWQ](https://github.com/casper-hansen/AutoAWQ.git)
|
| 15 |
+
```bash
|
| 16 |
+
pip install autoawq
|
| 17 |
+
```
|
| 18 |
+
### Inference with
|
| 19 |
+
```python
|
| 20 |
+
from awq import AutoAWQForCausalLM
|
| 21 |
+
import torch
|
| 22 |
+
from transformers import AutoTokenizer
|
| 23 |
+
|
| 24 |
+
prompt = "北京有什么好玩的地方?"
|
| 25 |
+
quant_path = "MiniCPM4.1-8B-AutoAWQ"
|
| 26 |
+
|
| 27 |
+
messages = [{"role": "user", "content": prompt}]
|
| 28 |
+
|
| 29 |
+
model = AutoAWQForCausalLM.from_quantized(
|
| 30 |
+
quant_path,
|
| 31 |
+
fuse_layers=False,
|
| 32 |
+
trust_remote_code=True
|
| 33 |
+
)
|
| 34 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
| 35 |
+
quant_path,
|
| 36 |
+
trust_remote_code=True
|
| 37 |
+
)
|
| 38 |
+
device = next(model.model.parameters()).device
|
| 39 |
+
|
| 40 |
+
# if enable_think
|
| 41 |
+
# formatted_prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt = True, enable_thinking = True)
|
| 42 |
+
# if disable_think
|
| 43 |
+
formatted_prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt = True, enable_thinking = False)
|
| 44 |
+
|
| 45 |
+
input_ids = tokenizer.encode(formatted_prompt, return_tensors='pt').to(device)
|
| 46 |
+
outputs = model.generate(
|
| 47 |
+
input_ids,
|
| 48 |
+
max_new_tokens=1000,
|
| 49 |
+
do_sample=True
|
| 50 |
+
)
|
| 51 |
+
# if enable think
|
| 52 |
+
# ans = [i.split("<|im_start|> assistant\n", 1)[1].strip() for i in tokenizer.batch_decode(outputs)]
|
| 53 |
+
# if disable think
|
| 54 |
+
ans = [i.split("<|im_start|> assistant\n<think>\n\n</think>", 1)[1].strip() for i in tokenizer.batch_decode(outputs)]
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
<p align="center">
|
| 58 |
+
<a href="https://github.com/OpenBMB/MiniCPM/" target="_blank">GitHub Repo</a> |
|
| 59 |
+
<a href="https://arxiv.org/abs/2506.07900" target="_blank">Technical Report</a> |
|
| 60 |
+
<a href="https://mp.weixin.qq.com/s/KIhH2nCURBXuFXAtYRpuXg?poc_token=HBIsUWijxino8oJ5s6HcjcfXFRi0Xj2LJlxPYD9c">Join Us</a>
|
| 61 |
+
</p>
|
| 62 |
+
<p align="center">
|
| 63 |
+
👋 Contact us in <a href="https://discord.gg/3cGQn9b3YM" target="_blank">Discord</a> and <a href="https://github.com/OpenBMB/MiniCPM/blob/main/assets/wechat.jpg" target="_blank">WeChat</a>
|
| 64 |
+
</p>
|