Repository Public Research preview
1{2 "name": "HTR-Tiny",3 "text_meta_prefix": false,4 "style_classes": null,5 "run": "medium_m1_tiny_mediumdenoiseinit_s3",6 "version": "0.3.0",7 "exported_at": "2026-11-10T06:45:15Z",8 "arch": "m1",9 "n_params": 33538945,10 "source_checkpoint": "artifacts/runs/medium_m1_tiny_mediumdenoiseinit_s3/best",11 "checkpoint_meta": {12 "created_at": "2026-11-10T03:32:58Z",13 "git": {14 "commit": "7d5e17e5a5c35df6229f5361adf79341ed2fdc13",15 "branch": "main",16 "dirty": false17 },18 "libraries": {19 "python": "3.14.4",20 "torch": "2.12.0",21 "numpy": "2.5.4",22 "anthropic": "2.3.0",23 "transformers": "5.8.1",24 "sentencepiece": "0.2.2",25 "lxml": "6.1.3",26 "pyarrow": "24.0.0",27 "sacrebleu": "2.6.0",28 "datasketch": "2.0.0",29 "rapidfuzz": "3.14.6"30 },31 "hardware": {32 "platform": "macOS-27.0-arm64-arm-64bit-Mach-O",33 "machine": "arm64",34 "cpu_count": 18,35 "chip": "Apple M5 Max",36 "ram_gb": 48.0,37 "torch_mps": true,38 "torch_cuda": false39 },40 "step": 16000,41 "tokens_seen": 72762134,42 "accel_seconds": 4450.0,43 "train_config": {44 "lr": 0.0002,45 "weight_decay": 0.01,46 "betas": [47 0.9,48 0.9849 ],50 "warmup_steps": 300,51 "max_steps": 20000,52 "accum": 1,53 "clip": 1.0,54 "eval_every": 1000,55 "log_every": 25,56 "patience": 5,57 "amp": "auto",58 "seed": 3,59 "max_tokens": 3072,60 "max_batch": 24,61 "loss": {62 "ce": 1.0,63 "fact": 0.0,64 "style": 0.0,65 "adv": 0.0,66 "con": 0.0,67 "sem": 0.0,68 "orth": 0.0,69 "label_smoothing": 0.1,70 "temperature": 0.0771 },72 "device": "auto",73 "keep_last": 274 },75 "eval": {76 "val_ce": 1.4702796535788918,77 "val_ppl": 4.350451590319647,78 "val_tokens": 184381279 }80 },81 "metrics": {82 "chrf": {83 "mean": 68.8415,84 "ci": [85 68.0331,86 69.614987 ]88 },89 "edit_sim": {90 "mean": 0.6786,91 "ci": [92 0.6706,93 0.686394 ]95 },96 "rougeL": {97 "mean": 0.6875,98 "ci": [99 0.6789,100 0.696101 ]102 },103 "exact_match": {104 "mean": 0.0,105 "ci": [106 0.0,107 0.0108 ]109 },110 "number_recall": {111 "mean": 0.9906,112 "ci": [113 0.9877,114 0.9934115 ]116 },117 "placeholders_exact": {118 "mean": 0.9953,119 "ci": [120 0.9923,121 0.9979122 ]123 },124 "_baseline_B0_identity": {125 "chrf": 68.4233,126 "edit_sim": 0.6785,127 "rougeL": 0.6841,128 "exact_match": 0.0,129 "number_recall": 1.0,130 "placeholders_exact": 1.0131 }132 },133 "training_data": "Medium stage: denoising pretraining on 258 264 rights-verified pre-2018 passages (PMC OA, CC BY economics/finance journals, arXiv, public-domain books; 20 000 steps), then fine-tuning on 72 698 validated (Haiku 5.5 rewrite → original) pairs of 14 000 passages; 32k SentencePiece; seed 3 (M* by the frozen validation rule). Validation metrics below are exploratory; confirmatory test results in the paper."134}135