{ "schema_version": "ingestion_model_v1", "backbone": "answerdotai/ModernBERT-large", "backbone_type": "modernbert", "weights_dtype": "bf16", "hidden_size": 1024, "max_position_embeddings": 8192, "vocab_size": 50368, "pad_token_id": 50283, "cls_token_id": 50281, "sep_token_id": 50282, "trained_args": { "checkpoint": "models/phase3/modernbert-large-mh-span-v3/best.pt", "train_file": "data/bakeoff/splits_v3/span_slot.train.jsonl", "val_file": "data/bakeoff/splits_v3/span_slot.val.jsonl", "out_dir": "models/phase4/modernbert-large-ldam-head", "summary": "runs/phase4/ldam-head-summary.json", "max_steps": 500, "drw_start": 400, "batch_size": 32, "lr": 0.0001, "warmup_steps": 50, "max_m": 0.5, "cb_beta": 0.9999, "seed": 42 }, "trained_steps": 500, "eval_at_save": {}, "heads": { "claim_present": { "kind": "linear", "out": 1, "pool": "cls" }, "predicate": { "kind": "linear", "out": 21, "pool": "cls" }, "subject_type": { "kind": "linear", "out": 13, "pool": "cls" }, "confidence": { "kind": "linear", "out": 1, "pool": "cls", "activation": "sigmoid" }, "span": { "kind": "linear", "out": 2, "pool": "per_token", "interpretation": "start/end logits; softmax over seq" } }, "role_prefix": "required: '[USER] ', '[ASSISTANT] ', '[TOOL_RESULT] ' prepended to blurb text before tokenization", "max_length": 512 }