|
--- |
|
language: |
|
- en |
|
datasets: |
|
- AIVision360 |
|
tags: |
|
- summarization |
|
- classification |
|
- translation |
|
- NLP |
|
- media and journalism |
|
- domain specific llm |
|
license: apache-2.0 |
|
pipeline_tag: text-generation |
|
--- |
|
|
|
# Llama2-7B-AIVision360 (News Connect) |
|
|
|
NewsConnect 7B (Llama2-7B-AIVision360) is a state-of-the-art, open-source chat model that stands as a beacon for technology, media, and AI news discussions. Built on the robust Llama2-7B architecture, this model has been enhanced and refined utilizing the AIVision360-8k dataset, making it a pioneer in the domain of AI news generation and interpretation. |
|
|
|
## Model Details |
|
|
|
- Architecture: Llama2-7B |
|
- Training Dataset: [AIVision360-8k](https://huggingface.co/datasets/ceadar-ie/AIVision360-8k) |
|
|
|
## Dataset Utilized: AIVision360-8k |
|
|
|
Drawing strength from the AIVision360-8k dataset, a curated collection hailing from "ainewshub.ie", this model is tailor-made for technology media and journalism. Offering structured interactions related to AI news, it captures the essence of the latest AI trends and evolutions. For a deeper dive into the dataset visit: [AIVision360-8k](https://huggingface.co/datasets/ceadar-ie/AIVision360-8k) |
|
|
|
### Model Specification |
|
|
|
- **Developed by:** CeADAR Connect Group |
|
- **Model type:** Large Language Model |
|
- **Language(s):** en |
|
- **Finetuned from model:** Llama2-7B |
|
|
|
## Key Features and Functionalities |
|
|
|
### Domain Specialization |
|
The Llama2-7B-AIVision360 model is specialized in AI news, serving as a resource for AI researchers, enthusiasts, and media experts. |
|
### Model API Accessibility |
|
Offers a straightforward Python integration for generating AI news insights. |
|
### Performance Optimisation |
|
Efficient performance across both CPU and GPU platforms. |
|
### Data Representation |
|
Utilises a comprehensive AI news dataset, enabling content generation to professional journalism standards. |
|
|
|
## Model Usage |
|
|
|
Experience the capabilities of the Llama2-7B-AIVision360 model through a well-structured Python interface. To kick-start your exploration, follow the steps and snippets given below: |
|
|
|
## Prerequisites |
|
### 1. Ensure required packages are available |
|
|
|
```python |
|
import torch |
|
import transformers |
|
from typing import Any, Dict |
|
from transformers import PreTrainedTokenizerFast, AutoTokenizer, |
|
AutoModelForCausalLM |
|
from transformers import ( |
|
AutoModelForCausalLM, |
|
AutoTokenizer, |
|
BitsAndBytesConfig, |
|
HfArgumentParser, |
|
TrainingArguments, |
|
pipeline, |
|
logging, |
|
) |
|
import time |
|
``` |
|
|
|
### 2. Initiate the model and tokenizer |
|
|
|
```python |
|
model_name = "ceadar-ie/Llama2-7B-AIVision360" |
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", |
|
torch_dtype=torch.float16, trust_remote_code=True) |
|
``` |
|
|
|
### 3. Create a function for generating text |
|
|
|
```python |
|
def generate_text(input_text): |
|
num_words_input = len(input_text.split()) |
|
|
|
start_time = time.time() |
|
|
|
pipe = pipeline(task="text-generation", model=model, tokenizer=tokenizer, max_length=500,temperature=1, top_k=10, top_p=0.1, repetition_penalty = 1.15) |
|
prompt = f" [INST] {input_text} [/INST]" |
|
result = pipe(prompt) |
|
generated_text = result[0]['generated_text'].replace(prompt, "").strip() |
|
|
|
end_time = time.time() |
|
time_taken = end_time - start_time |
|
|
|
num_words_generated = len(generated_text.split()) |
|
|
|
print("Generated Output:") |
|
print(generated_text) |
|
print('=' * 60 ) |
|
print("\nNumber of Words in Input Text:", num_words_input) |
|
print("Number of Words in Generated Output:", num_words_generated) |
|
print("Time Taken to Generate Output: {:.2f} seconds".format(time_taken)) |
|
``` |
|
|
|
## Example Evaluation and Use |
|
|
|
### Example Prompt 1: |
|
Concisely list any recent controversies or debates in the AI community. |
|
|
|
### Generated Output: |
|
The text discusses the latest news and developments in the AI industry, |
|
specifically highlighting controversies and debates that have been ongoing in the community. |
|
Recent controversies or debates in the AI community include: |
|
- Debate over whether AI-generated art is art. |
|
- Controversy over a study claiming that AIs can defeat humans at chess. |
|
- Disagreement among experts about the risk of extinction from AIs. |
|
- Concerns about the use of AI to manipulate images of children. |
|
- Proposals for new rules governing generative AI. |
|
- Criticism of companies using AI-powered customer service technology without proper disclosure. |
|
- Worries about the potential misuse of large language models like GPT-4. |
|
|
|
|
|
### Example Prompt 2: |
|
List new regulations or guidelines that have been introduced globally concerning AI ethics and bias mitigation. |
|
|
|
### Generated Output: |
|
There has been an increased focus on regulating artificial intelligence (AI) |
|
globally due to concerns about ethics, bias, and the potential for disinformation. New regulations |
|
or guidelines include: |
|
- European Commission's proposal for a law to ensure that AI systems are not biased against people with disabilities, race, gender, age, or religion. |
|
- OECD's recommendation on non-discrimination in AI decision-making. |
|
- ILO's updated recommendations on equal pay and other forms of discrimination. |
|
- Adoption of the Global Factorial Code by the UK government. |
|
- Benchmarks by the World Benchmark Foundation to measure the respect of human rights in AI products. |
|
- Tool by the Human Rights Foundation to assess respect of individual freedoms in generative AI applications. |
|
|
|
|
|
### Example Prompt 3: |
|
What are the current market trends in the AI industry? |
|
|
|
### Generated Output: |
|
The AI industry is rapidly evolving with notable trends such as: |
|
- Increased Adoption of Cloud-based AI Services. |
|
- Growing Focus on Explainability and Interpretability. |
|
- Integration of AI with Other Technologies. |
|
- Increased Focus on Ethics and Fairness. |
|
- Growing Demand for AI Talent. |
|
- Increased Focus on Cybersecurity |
|
|
|
|
|
### Example Prompt 4: |
|
Which sectors or industries are seeing the most rapid AI adoption right now? |
|
|
|
### Generated Output: |
|
AI adoption is evident across various sectors, including: |
|
- Healthcare: AI in analyzing medical images, diagnosing diseases, and chatbots for health advice. |
|
- Finance: Detecting fraud, analyzing financial data, and AI-powered financial advice. |
|
- Retail: Personalizing customer experiences, optimizing inventory, and AI chatbots for product recommendations. |
|
- Manufacturing: Optimizing production, predicting maintenance, and monitoring industrial equipment. |
|
- Transportation: Developing autonomous vehicles, improving traffic flow, and optimizing logistics. |
|
- Energy: Optimizing energy consumption, predicting demand, and enhancing renewable energy efficiency. |
|
- Agriculture: Optimizing yields, predicting weather, and monitoring agricultural equipment. |
|
- Education: Personalizing learning, grading assignments, and developing adaptive learning systems. |
|
- Security: Detecting cyber attacks, analyzing surveillance footage, and improving emergency responses. |
|
|
|
## Training Details |
|
|
|
### Training Hyperparameters |
|
- per_device_train_batch_size = 10 |
|
- gradient_accumulation_steps = 4 |
|
- optim = "paged_adamw_32bit" |
|
- warmup_steps = 100 |
|
- learning_rate = 2e-4 |
|
- max_grad_norm = 0.3 |
|
- warmup_ratio = 0.03 |
|
|
|
## Model Limitations |
|
Potential Biases: With its fine-tuning centered on AI news sources, inherent biases from these sources may reflect in the model's outputs. |
|
|
|
## Licensing |
|
The Llama2-7B-AIVision360 model, developed in collaboration with CeADAR Connect Group, combines the licensing frameworks of both Llama2 and AIVision360. Under Meta's terms, users are granted a non-exclusive, worldwide, non-transferable, royalty-free limited license for the use and modification of Llama Materials, inclusive of the Llama2 model and its associated documentation. When redistributing, the provided Agreement and a specific attribution notice must be included. In alignment with the AIVision360 dataset's licensing, the model is also distributed under the Apache 2.0 open-source license, promoting its use and modification within the AI community, while ensuring content reliability sourced from established AI news publishers. |
|
|
|
## Out-of-Scope Use |
|
Llama2-7B-AIVision360 is specifically tailored for AI news discussions. It is not optimized for: |
|
|
|
- General, non-AI-related conversations. |
|
- Domain-specific tasks outside AI news. |
|
- Direct interfacing with physical devices or applications. |
|
|
|
|
|
## Bias, Risks, and Limitations |
|
- Dataset Biases: The AIVision360-8k dataset may contain inherent biases that influence the model's outputs. |
|
- Over-reliance: The model is an aid, not a replacement for human expertise. Decisions should be made with careful consideration. |
|
- Content Understanding: The model lacks human-like understanding and cannot judge the veracity of news. |
|
- Language Limitations: The model's primary language is English. Performance may decrease with other languages. |
|
- Knowledge Cut-off: The model may not be aware of events or trends post its last training update. |
|
|
|
## Citation: |
|
``` |
|
@misc {ceadar_2023, |
|
author = { {CeADAR} }, |
|
title = { Llama2-7B-AIVision360 (Revision e349e9a) }, |
|
year = 2023, |
|
url = { https://huggingface.co/ceadar-ie/Llama2-7B-AIVision360 }, |
|
doi = { 10.57967/hf/1069 }, |
|
publisher = { Hugging Face } |
|
} |
|
``` |
|
## Contact: |
|
For any further inquiries or feedback concerning Llama2-7B-AIVision360, please forward your communications to ahtsham.zafar@ucd.ie |