File size: 2,848 Bytes
1f54b88
 
 
 
 
 
 
 
 
 
4946369
1f54b88
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3122c6c
51a604c
3122c6c
1f54b88
deefc28
ec8bb94
 
 
1f54b88
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4946369
da8efc4
 
255becb
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
---
pipeline_tag: text-generation
inference: true
widget:
- text: 'def factorial(n):'
  example_title: Factorial
  group: Python
- text: 'def recur_fibo(n):'
  example_title: Recursive Fibonacci
  group: Python
license: llama2
library_name: transformers
tags:
- text-generation
- code
language:
- en
---

# lemur-70b-v1

<p align="center">
  <img src="https://huggingface.co/datasets/OpenLemur/assets/resolve/main/lemur_icon.png" width="300" height="300" alt="Lemur">
</p>


<div align="center">
  <img src="https://huggingface.co/datasets/OpenLemur/assets/resolve/main/lemur_base_radar.png">
</div>

📄Paper: https://arxiv.org/abs/2310.06830

👩‍💻Code: https://github.com/OpenLemur/Lemur

## Use

### Setup

First, we have to install all the libraries listed in `requirements.txt` in [GitHub](https://github.com/OpenLemur/lemur-v1):

```bash
pip install -r requirements.txt
```

### Intended use

Since it is not trained on instruction following corpus, it won't respond well to questions like "What is the Python code to do quick sort?". 

### Generation

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("OpenLemur/lemur-70b-v1")
model = AutoModelForCausalLM.from_pretrained("OpenLemur/lemur-70b-v1", device_map="auto", load_in_8bit=True)

# Text Generation Example
prompt = "The world is "
input = tokenizer(prompt, return_tensors="pt")
output = model.generate(**input, max_length=50, num_return_sequences=1)
generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
print(generated_text)

# Code Generation Example
prompt = """
def factorial(n):
    if n == 0:
        return 1
"""
input = tokenizer(prompt, return_tensors="pt")
output = model.generate(**input, max_length=200, num_return_sequences=1)
generated_code = tokenizer.decode(output[0], skip_special_tokens=True)
print(generated_code)
```

# License
The model is licensed under the Llama-2 community license agreement.

# Acknowledgements
The Lemur project is an open collaborative research effort between [XLang Lab](https://www.xlang.ai/) and Salesforce Research. We thank Salesforce, Google Research and Amazon AWS for their gift support.
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_OpenLemur__lemur-70b-v1)

| Metric                | Value                     |
|-----------------------|---------------------------|
| Avg.                  | 54.03   |
| ARC (25-shot)         | 64.33          |
| HellaSwag (10-shot)   | 85.72    |
| MMLU (5-shot)         | 65.85         |
| TruthfulQA (0-shot)   | 44.78   |
| Winogrande (5-shot)   | 83.03   |
| GSM8K (5-shot)        | 28.73        |
| DROP (3-shot)         | 5.74         |