---
license: apache-2.0
---

```bash
python3 run_our_benchmark.py \
  WhirlwindAI/Arithmetic-SLM \
  --device cuda \
  --dtype bfloat16 \
  --batch-size auto \
  --skip-lm-eval \
  --our-benchmark-split train \
  --json-out bench_Benchmark_TEST.json
```

```bash
==========================================================================
OPEN SLM LEADERBOARD STYLE RESULT
==========================================================================
Model:  WhirlwindAI/Arithmetic-SLM
Params: 19,372,800
--------------------------------------------------------------------------
Benchmark          Task               Metric                          Score
--------------------------------------------------------------------------
Our Benchmark      WhirlwindAI/Benchm acc_custom_loglikelihood_mea   82.10%
--------------------------------------------------------------------------
Final Avg                                                               N/A
==========================================================================
our_benchmark details: 821/1000 correct, skipped=0, split=train

[INFO] Saved JSON: bench_Benchmark_TEST.json

```