arielcerdap/disfluency_roberta_large

Disfluency Labeling - Ariel Cerda

This model is a fine-tuned version of FacebookAI/roberta-large on the TimeStamped dataset. It achieves the following results on the evaluation set:

• Loss: 0.6249
• Precision: 0.0
• Recall: 0.0
• F1: 0.0
• Accuracy: 0.9075

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: 16
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.2995	0.2194	500	0.3309	0.9467	0.3143	0.4720	0.9365
0.354	0.4388	1000	0.2799	0.9079	0.3216	0.4750	0.9366
0.3717	0.6582	1500	0.2715	0.9458	0.3351	0.4948	0.9381
0.3362	0.8776	2000	0.3026	0.9153	0.3195	0.4737	0.9363
0.2668	1.0970	2500	0.3130	0.9519	0.3320	0.4923	0.9376
0.3311	1.3164	3000	0.2815	0.9687	0.3299	0.4922	0.9379
0.3345	1.5358	3500	0.3048	0.9976	0.2506	0.4006	0.9307
0.2945	1.7552	4000	0.2890	0.9621	0.3326	0.4943	0.9378
0.2648	1.9746	4500	0.2850	0.9740	0.3311	0.4942	0.9380
0.3272	2.1939	5000	0.2827	0.9657	0.3430	0.5062	0.9388
0.3161	2.4133	5500	0.2759	0.9237	0.3357	0.4924	0.9367
0.2687	2.6327	6000	0.2891	0.9757	0.3308	0.4941	0.9381
0.2948	2.8521	6500	0.2872	0.9784	0.3177	0.4796	0.9368
0.2608	3.0715	7000	0.2901	0.8284	0.3445	0.4866	0.9338
0.2947	3.2909	7500	0.2829	0.9572	0.3341	0.4954	0.9379
0.2939	3.5103	8000	0.2814	0.9702	0.3277	0.4900	0.9377
0.2581	3.7297	8500	0.2764	0.9757	0.3311	0.4944	0.9381
0.3108	3.9491	9000	0.2809	0.9721	0.3293	0.4919	0.9379
0.2929	4.1685	9500	0.2874	0.9737	0.3274	0.4901	0.9377
0.2939	4.3879	10000	0.2760	0.9689	0.3323	0.4949	0.9381
0.3173	4.6073	10500	0.2784	0.9722	0.3311	0.4940	0.9381
0.2784	4.8267	11000	0.2825	0.9709	0.3360	0.4992	0.9384
0.2593	5.0461	11500	0.2775	0.9724	0.3335	0.4967	0.9383
0.2507	5.2655	12000	0.2985	0.9708	0.3348	0.4978	0.9383
0.2707	5.4849	12500	0.2805	0.9714	0.3421	0.5060	0.9389
0.2775	5.7043	13000	0.2757	0.9697	0.3421	0.5057	0.9390
0.5178	5.9237	13500	0.4682	0.9052	0.0845	0.1545	0.9151
0.3553	6.1430	14000	0.3657	0.9574	0.1988	0.3292	0.9257
0.3496	6.3624	14500	0.3986	0.9565	0.1945	0.3233	0.9253
0.3452	6.5818	15000	0.4337	0.0	0.0	0.0	0.9075
0.3931	6.8012	15500	0.5834	0.0	0.0	0.0	0.9075
0.4035	7.0206	16000	0.5584	0.0	0.0	0.0	0.9075
0.3831	7.2400	16500	0.5585	0.0	0.0	0.0	0.9075
0.2817	7.4594	17000	0.5946	0.0	0.0	0.0	0.9075
0.3641	7.6788	17500	0.6069	0.0	0.0	0.0	0.9075
0.3866	7.8982	18000	0.6249	0.0	0.0	0.0	0.9075

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0