open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr1e-05_beta0.05_alpha1_epoch5

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_lr1e-05_beta0.05_alpha1_epoch5 model is a 1 billion parameter instruction-tuned language model, likely based on the Llama 3.2 architecture. This model is specifically designed for 'unlearning' tasks, indicated by 'unlearn_tofu' and 'forget10' in its name, suggesting it has been trained to remove specific information. It is optimized for scenarios requiring controlled forgetting or knowledge removal from a pre-trained LLM.

Loading preview...

Overview

This model, open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr1e-05_beta0.05_alpha1_epoch5, is a 1 billion parameter instruction-tuned language model. Its naming convention strongly suggests it is a Llama 3.2-based model that has undergone a specific 'unlearning' process, likely using the NPO (Neural Parameter Optimization) method, to forget approximately 10 specific pieces of information or data points. The model's parameters indicate a focus on efficient knowledge modification within a smaller LLM footprint.

Key Characteristics

  • Parameter Count: 1 billion parameters, offering a balance between performance and computational efficiency.
  • Instruction-Tuned: Designed to follow instructions effectively, making it suitable for various downstream tasks.
  • Unlearning Capability: The primary differentiator is its 'unlearning' feature, aimed at selectively removing specific knowledge or biases from the model.
  • Context Length: Supports a context length of 32768 tokens, allowing for processing longer inputs.

Potential Use Cases

  • Data Privacy: Useful for removing sensitive or private information from a model after training.
  • Bias Mitigation: Can be applied to reduce or eliminate unwanted biases learned during pre-training.
  • Controlled Knowledge Update: Enables targeted modification of a model's knowledge base without extensive retraining.
  • Research in Machine Unlearning: Serves as a valuable tool for exploring and developing techniques in machine unlearning.