Datasets:

---

WhirlwindAI

/

Benchmark-TEST

like 1

Follow

WhirlwindAI 10

@QuantaSparkLabs Now we can create our own benchmarks 🚀🤗, we just need to make a jsonl with "ctx": "question", "endings": [2 or 4 choices list str] and with "label": int (index of the list "endings" that is correct)

And we have run_our_benchmark.py, to test our models with lm_eval benchmarks or with our own benchmarks OUR_BENCHMARK_DATASET = "WhirlwindAI/Benchmark-TEST" ;)