CaGBERT-base_fl32_checkpoint-15852

Model Card for Model ID

We fine-tuned our base model for 71 epochs on the Ca dataset, epoch 62 showed the best macro average f1 score on the evaluation dataset.

Metrics

seqeval entity-wise in evaulate

eval_AVGf1 0.7889642398534424

eval_DIAGNOSIS.f1 0.7870941224825319

eval_DIAGNOSIS.precision 0.760222310440651

eval_DIAGNOSIS.recall 0.815935236472092

eval_DIAGNOSTIC.f1 0.7870518994114499

eval_DIAGNOSTIC.precision 0.7433046993431026

eval_DIAGNOSTIC.recall 0.8362706083001705

eval_DRUG.f1 0.9196581196581196

eval_DRUG.precision 0.8951747088186356

eval_DRUG.recall 0.945518453427065

eval_MEDICAL_FINDING.f1 0.7699975080986794

eval_MEDICAL_FINDING.precision 0.7438613384689456

eval_MEDICAL_FINDING.recall 0.7980371900826446

eval_THERAPY.f1 0.6810195496164316

eval_THERAPY.precision 0.64

eval_THERAPY.recall 0.7276573241671074

eval_accuracy 0.9332097564796261

eval_f1 0.7744305184135064

eval_loss 0.5050501823425293

eval_precision 0.7437801708132195

eval_recall 0.8077155722830835

eval_runtime 50.3125

eval_samples_per_second 162.624

eval_steps_per_second 20.333

test_AVGf1 0.7491200818619402

test_DIAGNOSIS.f1 0.703534151254349

test_DIAGNOSIS.precision 0.7192062897791089

test_DIAGNOSIS.recall 0.6885304659498208

test_DIAGNOSTIC.f1 0.7718579234972678

test_DIAGNOSTIC.precision 0.7573726541554959

test_DIAGNOSTIC.recall 0.786908077994429

test_DRUG.f1 0.9024472008045592

test_DRUG.precision 0.878016960208741

test_DRUG.recall 0.9282758620689655

test_MEDICAL_FINDING.f1 0.7280362842264404

test_MEDICAL_FINDING.precision 0.6848203939745076

test_MEDICAL_FINDING.recall 0.7770738704279225

test_THERAPY.f1 0.639724849527085

test_THERAPY.precision 0.6100861008610086

test_THERAPY.recall 0.6723904202440126

test_accuracy 0.9229989726085077

test_f1 0.7327920332701502

test_loss 0.6381183862686157

test_precision 0.7048546859693045

test_recall 0.7630354091792847

test_runtime 58.5022

test_samples_per_second 162.199

test_steps_per_second 20.29