In this assignment, we continued pretraining a LLaMA-3.2-1B model on Korean Domain using Unsloth with 4-bit quantization. We use the Korean subset of the [Wikipedia dataset] to continually pretrain the model. The codebase structure is based on the official practice notebook provided by Unsloth.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support