open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_IdkDPO_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_IdkDPO_lr2e-05_beta0.1_alpha2_epoch10 model is a 1 billion parameter instruction-tuned language model. This model is part of the Llama-3.2 family and has been specifically processed for unlearning, indicated by 'unlearn_tofu' and 'forget10'. It is likely optimized for scenarios requiring selective forgetting or knowledge removal from its training data, making it suitable for research in model unlearning and privacy-preserving AI.

Loading preview...

Model Overview

This model, unlearn_tofu_Llama-3.2-1B-Instruct_forget10_IdkDPO_lr2e-05_beta0.1_alpha2_epoch10, is a 1 billion parameter instruction-tuned variant from the Llama-3.2 family. Its naming convention suggests it has undergone a specific unlearning process, likely targeting the 'tofu' dataset with a 'forget10' configuration, using an IdkDPO (Implicit DPO) method with particular learning rate, beta, and alpha parameters over 10 epochs.

Key Characteristics

  • Parameter Count: 1 billion parameters, offering a balance between performance and computational efficiency.
  • Base Architecture: Derived from the Llama-3.2 series, indicating a strong foundation in general language understanding and generation.
  • Unlearning Focus: The model's primary differentiator is its explicit unlearning process, designed to remove specific information or biases from its knowledge base.
  • Instruction-Tuned: Capable of following instructions, making it versatile for various NLP tasks.

Potential Use Cases

  • Research in Machine Unlearning: Ideal for studying the effectiveness and mechanisms of unlearning algorithms.
  • Privacy-Preserving AI: Exploring methods to remove sensitive or unwanted information from models post-training.
  • Controlled Knowledge Models: Developing models where certain facts or concepts need to be selectively excluded or forgotten.
  • Comparative Analysis: Useful for comparing the behavior and performance of unlearned models against their original counterparts.