from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load model
model = AutoModelForCausalLM.from_pretrained(
    "aryan14072001/article-writer-rag",
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("aryan14072001/article-writer-rag")

# Generate article
messages = [
    {"role": "system", "content": "You are a professional journalist."},
    {"role": "user", "content": "Write article from:\n\n• Fact 1\n• Fact 2"}
]

prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.3)
article = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(article)

Training Details

Dataset: CNN/DailyMail articles
Method: LoRA (Low-Rank Adaptation)
Epochs: 5
Learning Rate: 5e-5
Anti-hallucination: Lower temperature, higher regularization

Limitations

Designed for news article generation
May not work well for other writing styles
Requires factual input notes

License

Apache 2.0