open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr2e-05_beta0.1_alpha2_epoch10

Hugging Face
TEXT GENERATIONPricing:Input $0.108 / Output $0.804Concurrent Unit Cost:1Model Size:1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:May 15, 2025Architecture:Transformer Featherless Exclusive Warm

The open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr2e-05_beta0.1_alpha2_epoch10 model is a 1 billion parameter instruction-tuned language model based on the Llama-3.2 architecture. This model is specifically designed for unlearning, having undergone a process to forget specific information. It is optimized for scenarios requiring controlled knowledge removal or privacy-preserving applications, offering a Llama-3.2-based foundation with targeted unlearning capabilities.

Loading preview...

Model Overview

This model, open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr2e-05_beta0.1_alpha2_epoch10, is a 1 billion parameter instruction-tuned language model built upon the Llama-3.2 architecture. Its primary distinguishing feature is its focus on unlearning, indicating it has been specifically processed to remove or forget certain information. This makes it distinct from standard instruction-tuned models, which typically aim to maximize knowledge retention.

Key Characteristics

  • Architecture: Based on the Llama-3.2 family, providing a strong foundation for general language understanding and generation.
  • Parameter Count: A compact 1 billion parameters, making it suitable for deployment in environments with resource constraints.
  • Instruction-Tuned: Capable of following instructions and performing various natural language tasks.
  • Unlearning Focus: The model name explicitly indicates a process of 'unlearning' (specifically 'forget10' and 'NPO' method), suggesting it has been modified to remove specific data or biases.

Potential Use Cases

  • Privacy-Preserving AI: Developing applications where certain sensitive information needs to be excluded from the model's knowledge base.
  • Controlled Knowledge: Creating models with intentionally limited or modified knowledge for specific, safe applications.
  • Research in Unlearning: Serving as a base for further research and experimentation in machine unlearning techniques and their impact on model performance and safety.
  • Resource-Efficient Deployment: Its 1B parameter size allows for easier deployment and inference compared to larger models, even with the specialized unlearning process.