{ "model": "Transformer (from scratch, Pre-LN, tied embeddings)", "task": "Pali → Sinhala MT", "vocab_size": 4000, "spm_model_type": "unigram", "d_model": 512, "n_heads": 8, "n_enc_layers": 5, "n_dec_layers": 5, "d_ff": 2048, "dropout": 0.3, "label_smoothing": 0.2, "beam_size": 8, "length_alpha": 0.6, "test_bleu": 22.9565, "test_chrf": 46.436, "checkpoint": "averaged", "val_bleu": 22.855 }