LinkedIn • Hugging Face • Medium •
simplifies the process of pruning LLMs using a convenient Colab notebook.
AutoPrune simplifies the process of pruning LLMs through Colab notebook. You need to specify pruning parameters as described in the pruning paper.
- Automated setup and execution using RunPod.
- Identifies the most ideal block of layers to prune
- Removes the pruning block and merges the rest of model
- Pushed to 🤗 Huggingface the pruned model.
- Doesn't perform the QLoRA-SFT (Healing) of the pruned model as it can be performed normally.