| [ |
| { |
| "scale_label": "1e17", |
| "best_cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.012, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.35 |
| }, |
| "best_value": 0.8514144989015371, |
| "n_trials": 12, |
| "all_trials": [ |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 8, |
| "matrix_lr": 0.03, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.9427201077211329, |
| "note": "baseline (warm-start from prior scale)" |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.03, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.8644734973437793, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 4, |
| "matrix_lr": 0.03, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.8796142617755859, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.01, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.855914794632743, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.003, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.9168695430147942, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.02, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.8546455830495197, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.015, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.8519772268389574, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.015, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.6 |
| }, |
| "value": 0.8555992013822545, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.015, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.3 |
| }, |
| "value": 0.852150737650948, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.015, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.2 |
| }, |
| "value": 0.8537460186091574, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.012, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.4 |
| }, |
| "value": 0.8514492010639352, |
| "note": null |
| }, |
| { |
| "cfg": { |
| "embedding_lr": 0.3, |
| "grad_clip": 1.0, |
| "depth": 6, |
| "matrix_lr": 0.012, |
| "warmup_ratio": 0.0, |
| "warmdown_ratio": 0.35 |
| }, |
| "value": 0.8514144989015371, |
| "note": null |
| } |
| ], |
| "save_path": "outputs/epiplexity_code_1e17_v0/tuner_1e17.json" |
| } |
| ] |