{ "pairs": "/data/datasets/monet-100k/train_pairs.jsonl", "val_pairs": "/data/datasets/monet-100k/validation_pairs.jsonl", "projector_checkpoint": "outputs/monet-100k/projector-siglip2-clean/projector.pt", "model": "/data/models/LFM-2.5-Encoder-230M", "vision_model": "google/siglip2-base-patch16-256", "output": "outputs/monet-100k/retrieval-matching-clean", "num_image_tokens": 32, "batch_size": 8, "steps": 1000, "max_text_length": 128, "lr": 5e-05, "temperature": 0.07, "matching_weight": 0.5, "lora_layers": 4, "lora_rank": 8, "lora_alpha": 16.0, "dtype": "float32", "seed": 42, "eval_every": 250, "eval_pairs": 256, "device": "cuda:0", "force": true }