File size: 6,737 Bytes
d32e815 3c7b2ea d32e815 3c7b2ea d32e815 3c7b2ea d32e815 3c7b2ea d32e815 3c7b2ea | 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 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 | ---
license: llama3
base_model: meta-llama/Meta-Llama-3-8B-Instruct
tags:
- RAG hallucinations
- Knowledge Conflict
datasets:
- qiaojin/PubMedQA
- Salesforce/FaithEval-counterfactual-v1.0
- wandb/RAGTruth-processed
metrics:
- accuracy
---
# CopyPasteLLM-L3-8B
## Quick Start
### Method 1: Using Transformers
```python
#!/usr/bin/env python3
import torch
from transformers import (
AutoTokenizer,
AutoModelForCausalLM,
BitsAndBytesConfig,
pipeline
)
import warnings
warnings.filterwarnings("ignore")
def load_model_and_tokenizer():
"""Load model and tokenizer"""
print("Loading model and tokenizer...")
# Base model configuration
base_model_name = "meta-llama/Meta-Llama-3-8B-Instruct"
lora_model_name = "wingchiuloong/CopyPasteLLM-L3-8B"
# Quantization configuration
quantization_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_compute_dtype=torch.float16,
bnb_4bit_use_double_quant=True,
bnb_4bit_quant_type="nf4"
)
# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(
base_model_name,
trust_remote_code=True
)
# Load base model
model = AutoModelForCausalLM.from_pretrained(
base_model_name,
quantization_config=quantization_config,
device_map="auto",
trust_remote_code=True,
torch_dtype=torch.float16
)
# Load LoRA weights
print("Loading LoRA weights...")
from peft import PeftModel
model = PeftModel.from_pretrained(model, lora_model_name)
return model, tokenizer
def create_pipeline(model, tokenizer):
"""Create inference pipeline"""
return pipeline(
"text-generation",
model=model,
tokenizer=tokenizer,
torch_dtype=torch.float16,
device_map="auto"
)
def format_prompt(user_input):
"""Format input to Llama-3 format"""
system_message = "You are a helpful AI assistant."
prompt = f"""<|begin_of_text|><|start_header_id|>system<|end_header_id|>
{system_message}<|eot_id|><|start_header_id|>user<|end_header_id|>
{user_input}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
"""
return prompt
def main():
"""Main function - One-time inference"""
print("CopyPasteLLM inference Demo")
print("=" * 50)
# Set test question
context = "Galileo Galilei, renowned as one of the most influential figures in the history of science, made numerous contributions that revolutionized our understanding of physics and astronomy. His meticulous work with telescopes led to groundbreaking discoveries about the moons of Jupiter and the phases of Venus. Beyond the realm of astronomy, his observations and experiments laid the foundation for classical mechanics. One of Galileo’s lesser-known achievements is his development of the Three Laws of Motion, which were critical in advancing the study of kinematics and dynamics. These laws articulate the principles of inertia, the relationship between force and motion, and the law of action and reaction, providing a comprehensive framework for understanding moving bodies. His work on pendulums also contributed substantially to timekeeping and horology, as he discovered that pendulums of different lengths oscillate at predictable periods, a principle still applied in modern clocks. Galileo’s interdisciplinary approach enabled him to synthesize knowledge from various fields, which allowed new theories to emerge, reshaping the scientific landscape of his time and beyond. Notably, his support of the heliocentric model of the solar system earned him both acclaim and censure, highlighting the tension between scientific inquiry and established doctrine. In contrast to Galen’s biological studies and Newton’s later contributions, Galileo’s articulation of the Three Laws of Motion was pivotal in the transition from Aristotelian physics to Newtonian mechanics. His contributions remain a testament to the interplay of observation, theory, and experimentation in scientific progress."
question = "Which law was Galileo Galilei responsible for describing?"
test_question = f"{context}\nQ: {question}\nA:"
print(f"{test_question}")
print("-" * 50)
try:
# Load model
model, tokenizer = load_model_and_tokenizer()
# Create pipeline
print("Creating inference pipeline...")
pipe = create_pipeline(model, tokenizer)
print("Model loaded! Starting inference...")
# Format input
prompt = format_prompt(test_question)
# Generate reply
print("Generating reply...")
outputs = pipe(
prompt,
max_new_tokens=512,
temperature=1.0,
top_p=0.95,
do_sample=True,
pad_token_id=tokenizer.eos_token_id,
eos_token_id=tokenizer.eos_token_id,
return_full_text=False
)
# Output result
response = outputs[0]['generated_text']
print(f"\nCopyPasteLLM:\n{response}") # According to the passage, Galileo Galilei was responsible for describing the Three Laws of Motion, which articulate the principles of inertia, the relationship between force and motion, and the law of action and reaction.
except Exception as e:
print(f"Inference failed: {e}")
print("Please ensure the necessary dependencies are installed: pip install transformers peft bitsandbytes accelerate")
if __name__ == "__main__":
main()
```
### Method 2: Using vLLM
```bash
base_model="meta-llama/Meta-Llama-3-8B-Instruct"
lora_modules_path="<model_at_your_huggingface_cache>" # like "~/.huggingface/hub/models--wingchiuloong--CopyPasteLLM-L3-8B/snapshots/<uuid>"
python -m vllm.entrypoints.openai.api_server \
--model $base_model \
--enable-lora \
--max-lora-rank 64 \
--lora-modules CopyPasteLLM-8b=$lora_modules_path \
--host 0.0.0.0 \
--port 8888 \
--gpu-memory-utilization 0.8 \
--max-model-len 1024 \
--max-num-seqs 32 \
--tensor-parallel-size 1
```
## Citation
If you use this model in your research, please cite:
```bibtex
Coming soon...
```
## License
This model is released under the same license as the base Llama-3-8B-Instruct model. Please refer to the [Llama 3 Community License](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) for details.
## Contact
For questions and support, please open an issue on the [GitHub repository](https://github.com/longyongchao/CopyPasteLLM).
---
**Note**: This model is designed for research and educational purposes. Please ensure responsible use and compliance with applicable laws and regulations. |