{ "final": { "optimizer": "adamw", "final_val_loss": 0.8029099716292618, "final_val_ppl": 2.2320266214687186, "final_test_loss": 0.8068099271516253, "final_test_ppl": 2.24074842228017, "final_test_bleu": 69.51636165889195, "total_tokens": 836214, "total_steps": 3731, "trainable_parameters": 33513984 }, "interval_metrics": [ { "tokens": 83757, "step": 373, "val_loss": 0.8545034069153569, "val_ppl": 2.350206994200092, "test_bleu": 67.52653732743481, "fraction": 0.1 }, { "tokens": 167424, "step": 746, "val_loss": 0.8446204476879458, "val_ppl": 2.3270943928712833, "test_bleu": 67.52653732743481, "fraction": 0.2 }, { "tokens": 251309, "step": 1119, "val_loss": 0.8286119679826285, "val_ppl": 2.290137748596516, "test_bleu": 67.52653732743481, "fraction": 0.3 }, { "tokens": 335115, "step": 1492, "val_loss": 0.822031588297234, "val_ppl": 2.2751172471807717, "test_bleu": 69.81881395594117, "fraction": 0.4 }, { "tokens": 418695, "step": 1866, "val_loss": 0.8211532175829391, "val_ppl": 2.273119728228765, "test_bleu": 73.8244075245152, "fraction": 0.5 }, { "tokens": 502198, "step": 2239, "val_loss": 0.8115048268871891, "val_ppl": 2.251293245530193, "test_bleu": 73.8244075245152, "fraction": 0.6 }, { "tokens": 585487, "step": 2612, "val_loss": 0.8100201664511, "val_ppl": 2.247953319460061, "test_bleu": 70.53499805805588, "fraction": 0.7 }, { "tokens": 668923, "step": 2985, "val_loss": 0.806964167611785, "val_ppl": 2.2410940630031315, "test_bleu": 69.81881395594117, "fraction": 0.8 }, { "tokens": 752480, "step": 3358, "val_loss": 0.8053326493114807, "val_ppl": 2.2374406581354904, "test_bleu": 67.52653732743481, "fraction": 0.9 }, { "tokens": 836214, "step": 3731, "val_loss": 0.8029099716292618, "val_ppl": 2.2320266214687186, "test_bleu": 73.20136104273496, "fraction": 1.0 } ] }