learning_rate: 5e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
num_epochs: 15
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Bleu	Ter	Meteor	Chrf
No log	1.0	81	1.9797	10.1144	73.2843	0.3715	37.0972
10.9084	2.0	162	1.5432	18.7172	63.3578	0.4670	45.9610
4.7877	3.0	243	1.3437	21.9419	59.9265	0.4994	50.1266
2.8577	4.0	324	1.2706	26.3645	56.3725	0.5270	51.5026
1.6373	5.0	405	1.2614	27.9727	54.4118	0.5490	54.1392
1.6373	6.0	486	1.2834	25.9852	58.2108	0.5171	52.8133
0.9539	7.0	567	1.3179	29.6546	55.2696	0.5405	55.7142
0.5791	8.0	648	1.2556	29.0924	53.1863	0.5463	55.3880
0.4108	9.0	729	1.2828	28.5718	55.1471	0.5428	54.9380
0.2697	10.0	810	1.3089	31.0715	53.5539	0.5570	56.3873
0.2697	11.0	891	1.3224	33.5406	50.3676	0.5706	57.3349
0.1766	12.0	972	1.3214	33.3119	51.3480	0.5692	57.1213
0.1218	13.0	1053	1.3342	33.1588	50.2451	0.5738	58.0382
0.0968	14.0	1134	1.3292	33.6328	49.8775	0.5809	58.4079
0.0758	15.0	1215	1.3335	32.8315	50.8578	0.5827	58.0866

josueu/m2m100-T4v1-es-zap-data

m2m100-T4v1-es-zap-data