PubMedBERT-mimic-phi-ner

This model is a fine-tuned version of microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0024
• F1 Macro: 0.9550
• F1 Weighted: 0.9550
• Precision: 0.9316
• Recall: 0.9795
• F1 Name: 0.95
• F1 Location: 0.93
• F1 Phone: 0.94
• F1 Date: 0.87
• F1 Mrn: 0.97
• F1 Account: 0.99
• F1 Age Over 89: 1.0
• F1 Device Id: 1.0
• F1 Ssn: 1.0
• F1 Url: 1.0
• F1 Email: 1.0

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Macro	F1 Weighted	Precision	Recall	F1 Name	F1 Location	F1 Phone	F1 Date	F1 Mrn	F1 Account	F1 Age Over 89	F1 Device Id	F1 Ssn	F1 Url	F1 Email
0.4473	0.1774	300	0.0939	0.4089	0.4089	0.2964	0.6588	0.58	0.26	0.38	0.17	0.35	0.2	0.06	0.4	0.17	0.0	0.05
0.0666	0.3547	600	0.0142	0.7808	0.7808	0.6804	0.9158	0.86	0.64	0.9	0.4	0.52	0.55	0.41	0.9	0.59	0.87	0.82
0.0225	0.5321	900	0.0070	0.8786	0.8786	0.8251	0.9394	0.89	0.79	0.94	0.44	0.84	0.98	0.9	0.95	0.98	1.0	0.99
0.0267	0.7094	1200	0.0049	0.91	0.91	0.8641	0.9610	0.92	0.84	0.76	0.66	0.95	0.99	0.95	0.99	1.0	1.0	1.0
0.0127	0.8868	1500	0.0046	0.9281	0.9281	0.8918	0.9675	0.93	0.87	0.94	0.78	0.96	0.99	1.0	1.0	1.0	1.0	1.0
0.0099	1.0638	1800	0.0035	0.9338	0.9338	0.9009	0.9691	0.93	0.91	0.94	0.84	0.96	0.99	1.0	1.0	0.99	1.0	1.0
0.0070	1.2412	2100	0.0030	0.9495	0.9495	0.9237	0.9768	0.95	0.9	0.95	0.88	0.96	0.99	1.0	1.0	1.0	1.0	1.0
0.0115	1.4186	2400	0.0026	0.9489	0.9489	0.9222	0.9771	0.94	0.93	0.95	0.86	0.97	0.99	1.0	1.0	1.0	1.0	1.0
0.0088	1.5959	2700	0.0025	0.9533	0.9533	0.9296	0.9784	0.95	0.92	0.95	0.88	0.96	0.99	1.0	1.0	1.0	1.0	1.0
0.0067	1.7733	3000	0.0024	0.9542	0.9542	0.9310	0.9787	0.95	0.92	0.94	0.87	0.97	0.99	1.0	1.0	1.0	1.0	1.0
0.0053	1.9506	3300	0.0024	0.9550	0.9550	0.9316	0.9795	0.95	0.93	0.94	0.87	0.97	0.99	1.0	1.0	1.0	1.0	1.0

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2