Adaptation of the test script for Alpaca Lora, plus 2 tests, Alpaca 7b and 13b

[AngainorDev]

• Test Script was adapted to handle alpaca lora only, from tloen repo https://github.com/tloen/alpaca-lora
  ex lora used: https://huggingface.co/Angainor/alpaca-lora-13b
• The params changed, call like
  python3 take_test.py --test_name='alpaca13b_lora_r16x4' --base_model='../LLAMA/llama-13b-hf' --lora_weights='../alpaca-lora/lora-alpaca' --trivia=hq_trivia_questions.json

This script variant is not meant to be merged, it's too different from the base, and handling different params for different models would make it more complex than needed. Better I think have several branches/forks.

• I did not touch the test files content.
  Tests were run with llama 7b and 13b, Alpaca lora rank 16.

For clarity sake, I insist on providing clear metrics in a single place in the script output.
Even after your answers, it's not clear to me what is a score or a count, and I have to manually compute stuff with number of answers * 2 to get the % of the table.
This does not help giving a proper reference to be used and clear understanding of the tests results.

Anyway, the raw result files are there if you want to add the correct metrics to the table.