{ "final": { "optimizer": "nadamw", "final_val_loss": 0.8026075718206293, "final_val_ppl": 2.2313517590897938, "final_test_loss": 0.8065095021121595, "final_test_ppl": 2.2400753464564516, "final_test_bleu": 69.06639392972643, "total_tokens": 836214, "total_steps": 3731, "trainable_parameters": 33513984 }, "interval_metrics": [ { "tokens": 83757, "step": 373, "val_loss": 0.8505439572895874, "val_ppl": 2.3409198660874333, "test_bleu": 70.53499805805588, "fraction": 0.1 }, { "tokens": 167424, "step": 746, "val_loss": 0.8383649248797544, "val_ppl": 2.3125826374506513, "test_bleu": 67.52653732743481, "fraction": 0.2 }, { "tokens": 251309, "step": 1119, "val_loss": 0.825183545370706, "val_ppl": 2.2822996324196647, "test_bleu": 67.52653732743481, "fraction": 0.3 }, { "tokens": 335115, "step": 1492, "val_loss": 0.8212574012155779, "val_ppl": 2.2733565624363847, "test_bleu": 69.81881395594117, "fraction": 0.4 }, { "tokens": 418695, "step": 1866, "val_loss": 0.8222601318223383, "val_ppl": 2.275637269918119, "test_bleu": 73.8244075245152, "fraction": 0.5 }, { "tokens": 502198, "step": 2239, "val_loss": 0.8109824475538101, "val_ppl": 2.2501175235785134, "test_bleu": 72.56198661222149, "fraction": 0.6 }, { "tokens": 585487, "step": 2612, "val_loss": 0.8097436169801303, "val_ppl": 2.247331735112173, "test_bleu": 67.52653732743481, "fraction": 0.7 }, { "tokens": 668923, "step": 2985, "val_loss": 0.8067046689236136, "val_ppl": 2.240512577484341, "test_bleu": 69.81881395594117, "fraction": 0.8 }, { "tokens": 752480, "step": 3358, "val_loss": 0.8050539144347423, "val_ppl": 2.236817092298246, "test_bleu": 67.52653732743481, "fraction": 0.9 }, { "tokens": 836214, "step": 3731, "val_loss": 0.8026075718206293, "val_ppl": 2.2313517590897938, "test_bleu": 71.23000806985263, "fraction": 1.0 } ] }