Update README.md
Browse files
README.md
CHANGED
|
@@ -7,7 +7,7 @@
|
|
| 7 |
<div align="center"><img src="misc/skywork_logo.jpeg" width="550"/></div>
|
| 8 |
|
| 9 |
<p align="center">
|
| 10 |
-
🤗 <a href="https://huggingface.co/Skywork" target="_blank">Hugging Face</a> • 🤖 <a href="https://modelscope.cn/organization/Skywork" target="_blank">ModelScope</a> • 👾 <a href="https://wisemodel.cn/organization/Skywork" target="_blank">Wisemodel</a> • 💬 <a href="https://github.com/SkyworkAI/Skywork/blob/main/misc/wechat.png?raw=true" target="_blank">WeChat</a>• 📜<a href="
|
| 11 |
</p>
|
| 12 |
|
| 13 |
<div align="center">
|
|
@@ -45,44 +45,13 @@ Skywork-MoE demonstrates comparable or superior performance to models with more
|
|
| 45 |
|
| 46 |
| | HuggingFace Model | ModelScope Model | Wisemodel Model |
|
| 47 |
|:-------:|:-----------:|:-----------------------------:|:-----------------------------:|
|
| 48 |
-
| **Skywork-MoE-base** | 🤗 [Skywork-MoE-base](https://
|
| 49 |
-
| **Skywork-MoE-Base-FP8** | 🤗 [Skywork-MoE-Base-FP8](https://
|
| 50 |
|
| 51 |
# Benchmark Results
|
| 52 |
We evaluated Skywork-MoE-base model on various popular benchmarks, including C-Eval, MMLU, CMMLU, GSM8K, MATH and HumanEval.
|
| 53 |
<img src="misc/skywork_moe_base_evaluation.png" alt="Image" width="600" height="280">
|
| 54 |
|
| 55 |
-
# Demonstration of Hugging Face Model Inference
|
| 56 |
-
|
| 57 |
-
## Base Model Inference
|
| 58 |
-
|
| 59 |
-
We can perform inference for the Skywork-MoE-base (16x13B size) model using HuggingFace on 8xA100/A800 or higher GPU hardware configurations.
|
| 60 |
-
|
| 61 |
-
```python
|
| 62 |
-
|
| 63 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 64 |
-
|
| 65 |
-
model = AutoModelForCausalLM.from_pretrained("Skywork/Skywork-MoE-base", trust_remote_code=True, device_map='auto')
|
| 66 |
-
tokenizer = AutoTokenizer.from_pretrained("Skywork/Skywork-MoE-base", trust_remote_code=True)
|
| 67 |
-
|
| 68 |
-
inputs = tokenizer('陕西的省会是西安', return_tensors='pt').to(model.device)
|
| 69 |
-
response = model.generate(inputs.input_ids, max_length=128)
|
| 70 |
-
print(tokenizer.decode(response.cpu()[0], skip_special_tokens=True))
|
| 71 |
-
"""
|
| 72 |
-
陕西的省会是西安。
|
| 73 |
-
西安,古称长安、镐京,是陕西省会、副省级市、关中平原城市群核心城市、丝绸之路起点城市、“一带一路”核心区、中国西部地区重要的中心城市,国家重要的科研、教育、工业基地。
|
| 74 |
-
西安是中国四大古都之一,联合国科教文组织于1981年确定的“世界历史名城”,美媒评选的世界十大古都之一。地处关中平原中部,北濒渭河,南依秦岭,八水润长安。下辖11区2县并代管西
|
| 75 |
-
"""
|
| 76 |
-
|
| 77 |
-
inputs = tokenizer('陕西的省会是西安,甘肃的省会是兰州,河南的省会是郑州', return_tensors='pt').to(model.device)
|
| 78 |
-
response = model.generate(inputs.input_ids, max_length=128)
|
| 79 |
-
print(tokenizer.decode(response.cpu()[0], skip_special_tokens=True))
|
| 80 |
-
"""
|
| 81 |
-
陕西的省会是西安,甘肃的省会是兰州,河南的省会是郑州,湖北的省会是武汉,湖南的省会是长沙,安徽的省会是合肥,江西的省会是南昌,江苏的省会是南京,浙江的省会是杭州,福建的省会是福州,广东的省会是广州,广西的省会是南宁,四川的省会是成都,贵州的省会是贵阳,云南的省会是昆明,山西的省会是太原,山东的省会是济南,河北的省会是石家庄,辽宁的省会是沈阳,吉林的省会是长春,黑龙江的
|
| 82 |
-
"""
|
| 83 |
-
|
| 84 |
-
```
|
| 85 |
-
|
| 86 |
|
| 87 |
# Demonstration of vLLM Model Inference
|
| 88 |
|
|
|
|
| 7 |
<div align="center"><img src="misc/skywork_logo.jpeg" width="550"/></div>
|
| 8 |
|
| 9 |
<p align="center">
|
| 10 |
+
🤗 <a href="https://huggingface.co/Skywork" target="_blank">Hugging Face</a> • 🤖 <a href="https://modelscope.cn/organization/Skywork" target="_blank">ModelScope</a> • 👾 <a href="https://wisemodel.cn/organization/Skywork" target="_blank">Wisemodel</a> • 💬 <a href="https://github.com/SkyworkAI/Skywork/blob/main/misc/wechat.png?raw=true" target="_blank">WeChat</a>• 📜<a href="https://github.com/SkyworkAI/Skywork-MoE/blob/main/skywork-moe-tech-report.pdf" target="_blank">Tech Report</a>
|
| 11 |
</p>
|
| 12 |
|
| 13 |
<div align="center">
|
|
|
|
| 45 |
|
| 46 |
| | HuggingFace Model | ModelScope Model | Wisemodel Model |
|
| 47 |
|:-------:|:-----------:|:-----------------------------:|:-----------------------------:|
|
| 48 |
+
| **Skywork-MoE-base** | 🤗 [Skywork-MoE-base](https://github.com/SkyworkAI/Skywork-MoE) | 🤖[Skywork-MoE-base](https://www.modelscope.cn/models/skywork/Skywork-MoE-base) | 👾[Skywork-MoE-base](https://wisemodel.cn/models/Skywork/Skywork-MoE-base) |
|
| 49 |
+
| **Skywork-MoE-Base-FP8** | 🤗 [Skywork-MoE-Base-FP8](https://github.com/SkyworkAI/Skywork-MoE) | 🤖 | 👾 |
|
| 50 |
|
| 51 |
# Benchmark Results
|
| 52 |
We evaluated Skywork-MoE-base model on various popular benchmarks, including C-Eval, MMLU, CMMLU, GSM8K, MATH and HumanEval.
|
| 53 |
<img src="misc/skywork_moe_base_evaluation.png" alt="Image" width="600" height="280">
|
| 54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
|
| 56 |
# Demonstration of vLLM Model Inference
|
| 57 |
|