clinical_bert

This model is a fine-tuned version of emilyalsentzer/Bio_ClinicalBERT on PlanTL-GOB-ES/pharmaconer. It achieves the following results on the evaluation and test set:

• Validation Loss: 1.6020
• Test Loss: 1.6591

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0005
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• lr_scheduler_warmup_steps: 100
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.78	100	1.9485
No log	1.56	200	1.8681
No log	2.34	300	1.8152
No log	3.12	400	1.7886
1.9285	3.91	500	1.7309
1.9285	4.69	600	1.6810
1.9285	5.47	700	1.7065
1.9285	6.25	800	1.7067
1.9285	7.03	900	1.7312
1.6644	7.81	1000	1.7006
1.6644	8.59	1100	1.6736
1.6644	9.38	1200	1.6846
1.6644	10.16	1300	1.6621
1.6644	10.94	1400	1.6381
1.5247	11.72	1500	1.6281
1.5247	12.5	1600	1.6605
1.5247	13.28	1700	1.6770
1.5247	14.06	1800	1.6666
1.5247	14.84	1900	1.6620
1.4334	15.62	2000	1.6677
1.4334	16.41	2100	1.6311
1.4334	17.19	2200	1.6743
1.4334	17.97	2300	1.6586
1.4334	18.75	2400	1.6086
1.3423	19.53	2500	1.6229
1.3423	20.31	2600	1.6475
1.3423	21.09	2700	1.6388
1.3423	21.88	2800	1.6275
1.3423	22.66	2900	1.6372
1.2712	23.44	3000	1.6345
1.2712	24.22	3100	1.6442
1.2712	25.0	3200	1.6864
1.2712	25.78	3300	1.6139
1.2712	26.56	3400	1.6161
1.215	27.34	3500	1.6491
1.215	28.12	3600	1.6442
1.215	28.91	3700	1.6409
1.215	29.69	3800	1.6539
1.215	30.47	3900	1.6052
1.1652	31.25	4000	1.6459
1.1652	32.03	4100	1.6362
1.1652	32.81	4200	1.6413
1.1652	33.59	4300	1.6377
1.1652	34.38	4400	1.6344
1.1213	35.16	4500	1.6406
1.1213	35.94	4600	1.6113
1.1213	36.72	4700	1.6410
1.1213	37.5	4800	1.6378
1.1213	38.28	4900	1.6341
1.0939	39.06	5000	1.6020

Framework versions

• Transformers 4.33.2
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3