open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr2e-05_beta0.05_alpha5_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_beta005_alpha5_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, indicating it has undergone a process to remove or reduce specific information from its training data. Its primary differentiator lies in its unlearning capabilities, making it suitable for use cases requiring data privacy or content moderation by selectively forgetting information.

Loading preview...

Model Overview

This model, open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr2e-05_beta0.05_alpha5_epoch10, is a 1 billion parameter instruction-tuned language model. It is built upon the Llama-3.2 architecture and has a context length of 32768 tokens. The model's name indicates it has undergone an "unlearning" process, specifically targeting "tofu" data with a "forget10" ratio using the NPO method, a learning rate of 2e-05, beta of 0.05, alpha of 5, and trained for 10 epochs.

Key Characteristics

  • Unlearning Focus: This model's primary characteristic is its application of machine unlearning techniques, suggesting it has been modified to forget specific data points or patterns from its original training. This is indicated by unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO in its name.
  • Instruction-Tuned: As an "Instruct" model, it is fine-tuned to follow instructions and perform various natural language processing tasks based on user prompts.
  • Llama-3.2 Base: It leverages the foundational capabilities of the Llama-3.2 architecture, providing a strong base for language understanding and generation.

Potential Use Cases

This model is particularly relevant for scenarios where:

  • Data Privacy is Critical: Users need to ensure certain sensitive information is no longer retrievable or generated by the model.
  • Content Moderation: It could be applied to remove biases or undesirable content learned during initial training.
  • Research in Unlearning: It serves as a practical example and tool for researchers exploring machine unlearning methodologies and their effectiveness.

Finetunes
1 models