Usage
Script Description
The script performs the following steps:

Loads LoRA weights from the specified directory.
Performs SVD on the weights.
Normalizes the data.
Prepares FAISS indexes for the training data.
Classifies validation data using the FAISS indexes.
Evaluates the model's performance.

Command Line Arguments
--seed: Random seed for reproducibility.
--rank: Rank for SVD.
--checkpoint: Checkpoint number to load.
--model_path: Path to the model.
--type: Type of dataset (imagenet, imagenet_1000, or dream).

Example running the Script:
python dsire.py --seed 42 --rank 32 --checkpoint 1200 --model_path /path/to/model --type imagenet

