naotodev/mateo1

Mateo-1

Modelo de IA conversacional neutro e versátil. Treinado para ser educado, profissional e objetivo. Respostas diretas e claras em português brasileiro.

Personality

Neutra - educado e objetivo

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

# Load base model
base_model = AutoModelForCausalLM.from_pretrained("gpt2")
tokenizer = AutoTokenizer.from_pretrained("gpt2")

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "mateo1")

# Generate response
prompt = "<|system|>Modelo de IA conversacional neutro e versátil. Tre <|user|>Ola! <|assistant|>"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=150, temperature=0.7)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Training Details

• Base Model: gpt2
• Method: LoRA (Low-Rank Adaptation)
• Language: Portuguese (Brazilian)
• Framework: PyTorch + Transformers + PEFT

Training

• Epochs: 3
• Final Loss: 8.8620
• Training Examples: ~829 exemplos de conversação e conhecimento geral em português brasileiro

Intended Use

This model is designed for conversational AI applications in Brazilian Portuguese. It can be used for:

• Educational assistance
• General conversation
• Knowledge questions
• Learning support

Limitations

• Small model size may limit complex reasoning
• Responses are in Brazilian Portuguese only
• May generate inaccurate information - always verify important facts
• Not suitable for critical applications without human oversight

Citation

@misc{mateo-core-mateo1,
  title={Mateo Core: Mateo-1},
  author={Intelliski AI},
  year={2026},
  url={https://huggingface.co/intelliski/mateo1}
}