ARBERT-base-submeter-classifier

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

• Loss: 0.1113
• Accuracy: 0.9709
• Macro F1: 0.6021
• Weighted F1: 0.9654

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: 5e-05
• train_batch_size: 128
• eval_batch_size: 256
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 256
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1	Weighted F1
1.2986	0.0865	500	0.7567	0.7652	0.2987	0.7566
0.5963	0.1730	1000	0.4073	0.8766	0.4198	0.8677
0.3789	0.2594	1500	0.2946	0.9137	0.4782	0.9067
0.2838	0.3459	2000	0.2346	0.9315	0.5072	0.9263
0.2401	0.4324	2500	0.2036	0.9422	0.5251	0.9357
0.2168	0.5189	3000	0.1820	0.9480	0.5299	0.9418
0.1955	0.6054	3500	0.1703	0.9519	0.5376	0.9455
0.1782	0.6919	4000	0.1578	0.9557	0.5499	0.9496
0.1699	0.7783	4500	0.1534	0.9571	0.5466	0.9508
0.1622	0.8648	5000	0.1496	0.9585	0.5579	0.9523
0.1569	0.9513	5500	0.1450	0.9602	0.5472	0.9537
0.1422	1.0377	6000	0.1469	0.9593	0.5690	0.9533
0.1251	1.1242	6500	0.1369	0.9628	0.5801	0.9568
0.123	1.2107	7000	0.1349	0.9632	0.5686	0.9571
0.1208	1.2972	7500	0.1337	0.9639	0.5747	0.9577
0.1188	1.3836	8000	0.1307	0.9645	0.5725	0.9581
0.1201	1.4701	8500	0.1295	0.9651	0.5705	0.9588
0.118	1.5566	9000	0.1271	0.9657	0.5692	0.9594
0.1167	1.6431	9500	0.1259	0.9662	0.5795	0.9600
0.1133	1.7296	10000	0.1220	0.9671	0.5803	0.9609
0.1116	1.8161	10500	0.1210	0.9674	0.5835	0.9612
0.1103	1.9025	11000	0.1165	0.9685	0.5836	0.9623
0.1093	1.9890	11500	0.1172	0.9683	0.5835	0.9623
0.0899	2.0754	12000	0.1174	0.9689	0.5972	0.9632
0.0851	2.1619	12500	0.1171	0.9692	0.5858	0.9632
0.0862	2.2484	13000	0.1172	0.9688	0.5902	0.9635
0.0835	2.3349	13500	0.1162	0.9696	0.5970	0.9638
0.0854	2.4213	14000	0.1159	0.9694	0.5925	0.9639
0.0797	2.5078	14500	0.1157	0.9698	0.5934	0.9640
0.0819	2.5943	15000	0.1152	0.9698	0.6004	0.9643
0.0813	2.6808	15500	0.1135	0.9703	0.5964	0.9646
0.0792	2.7673	16000	0.1120	0.9706	0.6000	0.9649
0.0786	2.8538	16500	0.1116	0.9707	0.6010	0.9653
0.0789	2.9402	17000	0.1113	0.9709	0.6021	0.9654

Framework versions

• Transformers 4.57.6
• Pytorch 2.10.0+cu128
• Datasets 3.6.0
• Tokenizers 0.22.2