Enrich llama input and output files

causalNLP / cladder

We develop benchmarks and analysis tools to evaluate the causal reasoning abilities of LLMs.

MIT License

89 stars 15 forks source link

Enrich llama input and output files #7

Closed feradauto closed 5 months ago

feradauto commented 5 months ago

This PR adds identification info such as question_id, rung and truth to the llama input file. This additional information is also included in the output file of the run_llama.py and run_alpaca.py scripts.

With these modifications, the output of run_llama.py can be scored with the Scorer class