bert-covid

This model is a fine-tuned version of microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext on the covid_qa_deepset dataset. It achieves the following results on the evaluation set:

Model description

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

Training Loss	Epoch	Step	Validation Loss
5.474	0.04	5	4.3730
3.9933	0.09	10	3.2783
3.0206	0.13	15	2.0289
1.9741	0.18	20	1.3879
1.4351	0.22	25	1.1733
1.5916	0.26	30	1.1623
0.5383	0.31	35	1.1952
0.7776	0.35	40	1.1920
1.1785	0.39	45	1.1216
1.1334	0.44	50	1.0412
0.7445	0.48	55	1.0829
0.6512	0.53	60	1.0443
0.7516	0.57	65	1.0089
0.5953	0.61	70	0.9273
0.8589	0.66	75	0.8947
0.7561	0.7	80	0.9009
0.9561	0.75	85	0.9006
0.7731	0.79	90	0.8482
0.8269	0.83	95	0.8380
0.9884	0.88	100	0.8200
0.9187	0.92	105	0.8775
0.585	0.96	110	0.8499
0.6835	1.01	115	0.8314
0.6668	1.05	120	0.7491
0.5558	1.1	125	0.7154
0.4491	1.14	130	0.8212
1.0667	1.18	135	0.8477
0.4472	1.23	140	0.7636
0.6892	1.27	145	0.7493
0.66	1.32	150	0.6932
0.5044	1.36	155	0.7675
0.5329	1.4	160	0.7406
0.2223	1.45	165	0.8099
0.5495	1.49	170	0.8758
0.5534	1.54	175	0.8476
0.4962	1.58	180	0.7953
0.7477	1.62	185	0.7610
0.7293	1.67	190	0.8357
0.6205	1.71	195	0.7339
0.5687	1.75	200	0.6908
0.884	1.8	205	0.6706
0.5928	1.84	210	0.6546
0.3209	1.89	215	0.6505
0.7585	1.93	220	0.6486
0.8501	1.97	225	0.6272
0.1664	2.02	230	0.6211
0.4483	2.06	235	0.6550
0.3361	2.11	240	0.6604
0.3085	2.15	245	0.6520
0.2407	2.19	250	0.6695
0.3418	2.24	255	0.6687
0.3165	2.28	260	0.6730
0.5811	2.32	265	0.6546
0.3516	2.37	270	0.6579
0.3136	2.41	275	0.6688
0.2508	2.46	280	0.6921
0.3463	2.5	285	0.7124
0.3603	2.54	290	0.7160
0.4455	2.59	295	0.6995
0.5433	2.63	300	0.6919
0.3411	2.68	305	0.6898
0.6065	2.72	310	0.6922
0.6258	2.76	315	0.6955
0.283	2.81	320	0.7008
0.6233	2.85	325	0.6988
0.3899	2.89	330	0.6949
0.238	2.94	335	0.6916
0.3166	2.98	340	0.6900