File size: 1,315 Bytes
ad2fd34
7c372d2
5df075b
 
c69c8f9
3df32f3
c69c8f9
 
 
 
 
 
 
 
 
3e359ce
d7a5687
5ac00d7
d7a5687
 
 
 
 
3ea4789
d7a5687
3ea4789
d7a5687
 
 
 
d5f2846
d7a5687
 
 
 
 
 
 
 
 
 
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
---
license: openrail
model_creator: axiong
model_name: PMC_LLaMA_13B
---
# PMC_LLaMA_13B - AWQ
- Model creator: [axiong](https://huggingface.co/axiong)
- Original model: [PMC_LLaMA_13B](https://huggingface.co/axiong/PMC_LLaMA_13B)

## Description

This repo contains AWQ model files for [PMC_LLaMA_13B](https://huggingface.co/axiong/PMC_LLaMA_13B).

### About AWQ

[Activation-aware Weight Quantization (AWQ)](https://arxiv.org/abs/2306.00978) selectively preserves a subset of crucial weights for LLM performance instead of quantizing all weights in a model. This targeted approach minimizes quantization loss, allowing models to operate in 4-bit precision without compromising performance.

Example of usage with vLLM library:

```python
from vllm import LLM, SamplingParams

prompts = [
    "What is the mechanism of action of antibiotics?",
    "How do statins work to lower cholesterol levels?",
    "Tell me about Paracetamol"
]

sampling_params = SamplingParams(temperature=0.8)

llm = LLM(model="disi-unibo-nlp/pmc-llama-13b-awq", quantization="awq", dtype="half")

outputs = llm.generate(prompts, sampling_params)

# Print the outputs.
for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(f"Prompt: {prompt}")
    print(f"Response: {generated_text}")
```