mzzavaa/otto_model_demo_1

OTTO Model Demo 1 - LoRA Adapter

Model Details

Model Description

OTTO Model Demo 1 is a LoRA-adapted model fine-tuned on AWS-related queries, leveraging the InstructLab framework. The model is built on instructlab/merlinite-7b-lab and trained using Low-Rank Adaptation (LoRA) to enhance instruction-following capabilities, particularly for AWS User Group and cloud computing-related questions.

• Developed by: mzzavaa
• Model type: LoRA Adapter for Instruction-Tuned LLMs
• Language(s): English
• License: MIT
• Finetuned from model: instructlab/merlinite-7b-lab

Model Sources

• Repository: Hugging Face Model Page
• Base Model: instructlab/merlinite-7b-lab

Uses

Direct Use

This adapter is fine-tuned for AWS-related queries and instruction-based interactions, such as:

• AWS User Group-related questions
• Technical cloud computing inquiries
• Community-driven event and FAQ automation

Out-of-Scope Use

• General-purpose conversation outside of AWS topics
• Real-time AWS infrastructure decision-making

How to Get Started with the Model

Since this is a LoRA adapter, it must be loaded on top of the base model.

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

# Load base model
base_model = AutoModelForCausalLM.from_pretrained("instructlab/merlinite-7b-lab", torch_dtype="auto")

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "mzzavaa/otto_model_demo_1")

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained("instructlab/merlinite-7b-lab")

# Test inference
input_text = "What is the typical time and venue for AWS User Group Vienna meetups?"
inputs = tokenizer(input_text, return_tensors="pt")
output = model.generate(**inputs)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Training Details

Training Data

• Synthetic AWS-related queries generated using ilab data generate

Training Procedure

• Base Model: instructlab/merlinite-7b-lab
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Precision: bf16 for efficiency
• Optimizer: AdamW
• Learning Rate: 2e-5

Evaluation

Testing Data, Factors & Metrics

Metrics

Metric	Value
Accuracy	[To be tested]
Perplexity	[To be tested]
F1 Score	[To be tested]

Bias, Risks, and Limitations

Known Limitations

• This LoRA adapter only works well for AWS-related topics.
• Since it’s a lightweight adapter, it inherits biases from the base model.

Citation

If you use this model, please cite:

@misc{mzzavaa2024otto,
  title={OTTO Model Demo 1 - LoRA Adapter for AWS},
  author={mzzavaa},
  year={2025},
  howpublished={\url{https://huggingface.co/mzzavaa/otto_model_demo_1}}
}

Model Card Contact

For questions or feedback, reach out via Hugging Face or email at: [Your Contact Info]

Framework versions

• PEFT 0.14.0