GOBO Baseline script

This is a baseline implementation of GOBO Quantization.

This script quantizes weight matrices of a model in a numpy array format.

To use this script:

1. Extract the state dictionary of a model and store each weight matrix as a numpy array (.npy) .
2. Copy and run this script in the same directory as the weight matrices.
3. Replace the quantized weights in your model's state dictionary and run the model.