Adaptation of the test script for Alpaca Lora, plus 2 tests, Alpaca 7b and 13b #6
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
ex lora used: https://huggingface.co/Angainor/alpaca-lora-13b
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.
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.