Question Answering
PEFT
English
File size: 2,237 Bytes
3bd7bd6
 
51d6404
 
 
 
 
 
 
 
3bd7bd6
bb8ceb9
3bd7bd6
bb8ceb9
3bd7bd6
bb8ceb9
3bd7bd6
bb8ceb9
3bd7bd6
bb8ceb9
 
 
 
 
 
 
 
 
01ffd01
 
 
 
 
bb8ceb9
 
 
 
 
 
 
 
 
 
ac6dbde
 
bb8ceb9
01ffd01
bb8ceb9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
library_name: peft
license: mit
datasets:
- timdettmers/openassistant-guanaco
- tatsu-lab/alpaca
- BI55/MedText
language:
- en
pipeline_tag: question-answering
---
 Here is a README.md explaining how to run the Archimedes model locally:

# Archimedes Model

This README provides instructions for running the Archimedes conversational AI assistant locally.

## Requirements

- Python 3.6+
- [Transformers](https://huggingface.co/docs/transformers/installation) 
- [Peft](https://github.com/hazyresearch/peft)
- PyTorch 
- Access to the LLAMA 2 model files or a cloned public model

Install requirements:

```
!pip install transformers
!pip install peft
!pip install torch
!pip install datasets
!pip install bitsandbytes
```

## Usage

```python
import transformers
from peft import LoraConfig, get_peft_model  
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig

login() # Need access to the gated model.

# Load LLAMA 2 model
model_name = "meta-llama/Llama-2-7b-chat-hf" 

# Quantization configuration 
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.float16,
)

# Load model
model = AutoModelForCausalLM.from_pretrained(
    model_name, 
    quantization_config=bnb_config,
    trust_remote_code=True
)

# Load LoRA configuration
lora_config = LoraConfig.from_pretrained('harpyerr/archimedes-300s-7b-chat')
model = get_peft_model(model, lora_config) 

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
tokenizer.pad_token = tokenizer.eos_token

# Define prompt  
text = "Can you tell me who made Space-X?"
prompt = "You are a helpful assistant. Please provide an informative response. \n\n" + text

# Generate response
device = "cuda:0" 
inputs = tokenizer(prompt, return_tensors="pt").to(device)
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

This loads the LLAMA 2 model, applies 4-bit quantization and LoRA optimizations, constructs a prompt, and generates a response.

See the [docs](https://huggingface.co/docs/transformers/model_doc/auto#transformers.AutoModelForCausalLM) for more details.