This dataset contains the compressed model weights from tensor network compression methodology applied to GPT-2 XL.
📁 Files Included
Compressed Model Weights (.pt files)
compressed_gpt2_xl_68.3%.pt - Base compressed model (~68% compression)
compressed_gpt2_xl_68.3%_healed.pt - Compressed + knowledge distillation healing
compressed_gpt2_xl_68.3%_enwik8_trained.pt - Compressed + enwik8 fine-tuning… See the full description on the dataset page: https://huggingface.co/datasets/prompterminal/gpt2-compressed-weights.
