disfluency-4-deberta-v2

This model is a fine-tuned version of on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0117
• Model Preparation Time: 0.0032
• Accuracy: 0.9627
• Balanced Accuracy: 0.9358
• F1: 0.9037
• F1 Macro: 0.9403
• Precision: 0.9162
• Precision Macro: 0.9449
• Recall: 0.8917
• Recall Macro: 0.9358
• Specificity: 0.9800
• Sensitivity: 0.8917
• False Positive Rate: 0.0200
• False Negative Rate: 0.1083
• Mcc: 0.8807
• True Positives: 7999
• False Positives: 732
• False Negatives: 972
• True Negatives: 35943

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• 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_with_restarts
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 6

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Accuracy	Balanced Accuracy	F1	F1 Macro	Precision	Precision Macro	Recall	Recall Macro	Specificity	Sensitivity	False Positive Rate	False Negative Rate	Mcc	True Positives	False Positives	False Negatives	True Negatives
0.0245	0.2308	200	0.0237	0.0032	0.8388	0.8286	0.6644	0.7792	0.5624	0.7554	0.8118	0.8286	0.8455	0.8118	0.1545	0.1882	0.5794	7283	5668	1688	31007
0.0179	0.4616	400	0.0173	0.0032	0.9142	0.8860	0.7936	0.8697	0.7525	0.8560	0.8395	0.8860	0.9325	0.8395	0.0675	0.1605	0.7414	7531	2477	1440	34198
0.0153	0.6924	600	0.0149	0.0032	0.9362	0.9114	0.8428	0.9014	0.8167	0.8923	0.8707	0.9114	0.9522	0.8707	0.0478	0.1293	0.8035	7811	1753	1160	34922
0.0131	0.9233	800	0.0135	0.0032	0.9410	0.9208	0.8554	0.9092	0.8257	0.8988	0.8874	0.9208	0.9542	0.8874	0.0458	0.1126	0.8193	7961	1681	1010	34994
0.0121	1.1535	1000	0.0128	0.0032	0.9574	0.9264	0.8898	0.9317	0.9046	0.9372	0.8754	0.9264	0.9774	0.8754	0.0226	0.1246	0.8635	7853	828	1118	35847
0.0115	1.3843	1200	0.0120	0.0032	0.9549	0.9324	0.8864	0.9291	0.8777	0.9260	0.8952	0.9324	0.9695	0.8952	0.0305	0.1048	0.8583	8031	1119	940	35556
0.0102	1.6151	1400	0.0118	0.0032	0.9605	0.9341	0.8987	0.9371	0.9071	0.9402	0.8904	0.9341	0.9777	0.8904	0.0223	0.1096	0.8743	7988	818	983	35857
0.0102	1.8459	1600	0.0117	0.0032	0.9627	0.9358	0.9037	0.9403	0.9162	0.9449	0.8917	0.9358	0.9800	0.8917	0.0200	0.1083	0.8807	7999	732	972	35943
0.0085	2.0762	1800	0.0121	0.0032	0.9607	0.9386	0.9003	0.9379	0.8986	0.9373	0.9020	0.9386	0.9751	0.9020	0.0249	0.0980	0.8759	8092	913	879	35762
0.0092	2.3070	2000	0.0113	0.0032	0.9613	0.9400	0.9019	0.9389	0.8990	0.9379	0.9049	0.9400	0.9751	0.9049	0.0249	0.0951	0.8779	8118	912	853	35763
0.0085	2.5378	2200	0.0115	0.0032	0.9617	0.9411	0.9030	0.9396	0.8989	0.9381	0.9071	0.9411	0.9751	0.9071	0.0249	0.0929	0.8792	8138	915	833	35760
0.0083	2.7686	2400	0.0111	0.0032	0.9613	0.9418	0.9024	0.9391	0.8952	0.9365	0.9096	0.9418	0.9740	0.9096	0.0260	0.0904	0.8783	8160	955	811	35720
0.0083	2.9994	2600	0.0111	0.0032	0.9618	0.9421	0.9036	0.9399	0.8975	0.9377	0.9097	0.9421	0.9746	0.9097	0.0254	0.0903	0.8798	8161	932	810	35743

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.1