distilbert-classn-LinearAlg-finetuned-span-width-3

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

• Loss: 1.0465
• Accuracy: 0.7222
• F1: 0.7251
• Precision: 0.7350
• Recall: 0.7222

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• 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
• lr_scheduler_warmup_steps: 500
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
5.0331	0.6849	50	2.4889	0.0476	0.0310	0.0241	0.0476
4.8321	1.3699	100	2.4703	0.0794	0.0509	0.0396	0.0794
4.8811	2.0548	150	2.4370	0.0873	0.0587	0.0487	0.0873
4.8198	2.7397	200	2.4201	0.0952	0.0662	0.0670	0.0952
4.7571	3.4247	250	2.4151	0.1190	0.0876	0.1070	0.1190
4.6927	4.1096	300	2.3845	0.1270	0.1051	0.1038	0.1270
4.607	4.7945	350	2.3643	0.1508	0.1431	0.1732	0.1508
4.5543	5.4795	400	2.3641	0.1508	0.1376	0.1467	0.1508
4.2468	6.1644	450	2.2960	0.1984	0.1655	0.1800	0.1984
4.1548	6.8493	500	2.1901	0.2381	0.2339	0.3055	0.2381
3.7031	7.5342	550	2.0601	0.3571	0.3299	0.4862	0.3571
3.4466	8.2192	600	2.0129	0.3651	0.3678	0.4649	0.3651
3.0481	8.9041	650	1.8144	0.4365	0.4292	0.4630	0.4365
2.5507	9.5890	700	1.6802	0.4921	0.4820	0.4894	0.4921
2.1803	10.2740	750	1.5281	0.5635	0.5703	0.6068	0.5635
1.7031	10.9589	800	1.4110	0.5714	0.5538	0.5561	0.5714
1.4117	11.6438	850	1.3102	0.6349	0.6388	0.6649	0.6349
1.0765	12.3288	900	1.2092	0.6746	0.6693	0.6834	0.6746
0.8571	13.0137	950	1.2143	0.6746	0.6679	0.6859	0.6746
0.6671	13.6986	1000	1.1043	0.6905	0.6811	0.6961	0.6905
0.5448	14.3836	1050	1.0635	0.7063	0.7057	0.7239	0.7063
0.419	15.0685	1100	1.0836	0.7381	0.7366	0.7522	0.7381
0.3435	15.7534	1150	1.0320	0.7063	0.7130	0.7487	0.7063
0.2654	16.4384	1200	1.0282	0.7063	0.7026	0.7209	0.7063
0.1986	17.1233	1250	1.0172	0.7063	0.7076	0.7218	0.7063
0.1714	17.8082	1300	1.0305	0.7302	0.7297	0.7572	0.7302
0.118	18.4932	1350	1.0045	0.7302	0.7293	0.7456	0.7302
0.1293	19.1781	1400	1.0415	0.7381	0.7402	0.7566	0.7381
0.0934	19.8630	1450	1.0429	0.7143	0.7183	0.7376	0.7143
0.0598	20.5479	1500	1.0438	0.7302	0.7310	0.7397	0.7302
0.0651	21.2329	1550	1.0299	0.7143	0.7187	0.7335	0.7143
0.0618	21.9178	1600	1.0538	0.7143	0.7185	0.7313	0.7143
0.0664	22.6027	1650	1.0280	0.7381	0.7394	0.7552	0.7381
0.0662	23.2877	1700	1.0319	0.7302	0.7320	0.7426	0.7302
0.0315	23.9726	1750	1.0467	0.7222	0.7251	0.7350	0.7222
0.0462	24.6575	1800	1.0465	0.7222	0.7251	0.7350	0.7222

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0