| { |
| "model": "Sol Nano", |
| "parameters": 2895188, |
| "tokens_seen": 5000000000, |
| "checkpoint_sha256": "682584ee247f1dd9d9c2fd702a79e5713f5d9c2f61b78e8fe7fd08ec7054db5e", |
| "tokenizer_sha256": "72a97c4c592c991733fcc6ad687734ce51e1ba25b0f7c2faf611ac30edc50231", |
| "dtype": "float32", |
| "attention": "Triton compiled FlexAttention, no SDPA", |
| "max_length": 512, |
| "scored_requests": 62032, |
| "requests_requiring_context_truncation": 0, |
| "continuations_longer_than_context": 0, |
| "torch_version": "2.14.0+cu130", |
| "lm_eval_version": "0.4.12", |
| "sample_counts": { |
| "hellaswag": 10042, |
| "arc_easy": 2376, |
| "arc_challenge": 1172, |
| "piqa": 1838 |
| }, |
| "updates": 19074, |
| "checkpoint": "/marimo/sol_nano_5b_mixed_live/checkpoint.pt", |
| "scores_percent": { |
| "hellaswag": 28.40071698864768, |
| "arc_easy": 32.07070707070707, |
| "arc_challenge": 21.160409556313994, |
| "piqa": 53.91730141458106, |
| "arithmark3": 33.8 |
| }, |
| "open_slm_intelligence_index": 6.068393579554266, |
| "protocol": "zero-shot normalized accuracy; LM-evaluation-harness 0.4.12 plus official ArithMark-3.0", |
| "elapsed_seconds": 106.02102780342102 |
| } |
|
|
