compress_llama.sh

#!/bin/bash

# run data whitening with 20% compression ratio
python SVDLLM.py --model jeffwan/llama-7b-hf --step 1 --ratio 0.2 --whitening_nsamples 256 --dataset wikitext2 --seed 3 --model_seq_len 2048 --save_path .
## you can also run the following command for low-resource gpu (ex. llama 7b will only need 15G gpu memory to compress) or to compress large-scale llm (ex. llama 65b)
# python SVDLLM.py --model jeffwan/llama-7b-hf --step 1 --ratio 0.2 --whitening_nsamples 256 --dataset wikitext2 --model_seq_len 2048 --save_path ./ --run_low_resource

# evaluate the perplexity of llama_7b_whitening_0.2.pt
python SVDLLM.py --step 4 --model_path whitening/jeffwan_llama_7b_hf_whitening_0.2.pt
# evaluate the efficiency of llama_7b_whitening_0.2.pt
python SVDLLM.py --step 5 --model_path whitening/jeffwan_llama_7b_hf_whitening_0.2.pt