pgajo/mbert_EW-TT-PE_U1_S0_DROP1_mbert_E10_DEV98.0

Model description:

Model: bert-base-multilingual-cased

Dataset: TASTEset

Unshuffled ratio: ['1']

Shuffled ratio: ['0']

Best exact match epoch: 10

Best exact match: 97.79

Best epoch: 10

Drop duplicates: ['1']

Max epochs = 10

Optimizer lr = 3e-05

Optimizer eps = 1e-08

Batch size = 32

Dataset path = pgajo/EW-TT-PE_U1_S0_DROP1_mbert

Results

epoch	train_loss	train_f1	train_exact	dev_loss	dev_f1	dev_exact	test_loss	test_f1	test_exact
1	2.94	17.7	10.02	0.76	69.4	58.56	0	0	0
2	0.38	86.34	80.72	0.15	95.66	91.99	0	0	0
3	0.08	97.33	95.44	0.08	98.38	95.86	0	0	0
4	0.04	98.98	98.27	0.09	98.09	96.41	0	0	0
5	0.03	98.94	98.41	0.08	98.44	96.41	0	0	0
6	0.02	99.32	98.76	0.08	98.57	97.24	0	0	0
7	0.02	99.44	99.24	0.05	98.44	97.51	0	0	0
8	0.01	99.82	99.59	0.07	98.47	97.24	0	0	0
9	0.01	99.8	99.65	0.07	98.66	97.24	0	0	0
10	0.01	99.82	99.65	0.06	98.59	97.79	0	0	0