roberta-base-multi-head

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

• Loss: 0.4882
• Accuracy: 0.5566
• F1 Macro: 0.5333
• F1 Micro: 0.5566
• Precision Macro: 0.5431
• Recall Macro: 0.5389
• Roc Auc: 0.7826

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-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• 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_ratio: 0.05
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	F1 Micro	Precision Macro	Recall Macro	Roc Auc
No log	0.1304	200	0.6818	0.2672	0.1648	0.2672	0.1218	0.2567	0.4838
No log	0.2609	400	0.6261	0.3230	0.1221	0.3230	0.0808	0.25	0.5071
0.6589	0.3913	600	0.5625	0.3902	0.2186	0.3902	0.2036	0.2855	0.5948
0.6589	0.5217	800	0.5461	0.4307	0.2771	0.4307	0.3373	0.3294	0.6677
0.5528	0.6522	1000	0.5142	0.4806	0.3522	0.4806	0.4562	0.3832	0.7032
0.5528	0.7826	1200	0.5025	0.4966	0.3990	0.4966	0.4866	0.4231	0.7188
0.5528	0.9130	1400	0.5006	0.4939	0.4140	0.4939	0.4853	0.4429	0.7312
0.5111	1.0430	1600	0.4903	0.5165	0.4163	0.5165	0.5065	0.4369	0.7386
0.5111	1.1735	1800	0.4821	0.5267	0.4650	0.5267	0.5003	0.4699	0.7494
0.4847	1.3039	2000	0.4803	0.5273	0.4900	0.5273	0.5013	0.4970	0.7582
0.4847	1.4343	2200	0.4742	0.5438	0.5020	0.5438	0.5153	0.5015	0.7637
0.4847	1.5648	2400	0.4672	0.5476	0.4998	0.5476	0.5270	0.4976	0.7692
0.47	1.6952	2600	0.4743	0.5396	0.4820	0.5396	0.5346	0.4885	0.7650
0.47	1.8256	2800	0.4675	0.5512	0.5104	0.5512	0.5282	0.5029	0.7734
0.4651	1.9561	3000	0.4671	0.5436	0.5151	0.5436	0.5211	0.5190	0.7747
0.4651	2.0861	3200	0.4631	0.5643	0.5269	0.5643	0.5431	0.5209	0.7804
0.4651	2.2165	3400	0.4681	0.5445	0.5109	0.5445	0.5359	0.5207	0.7798
0.4415	2.3469	3600	0.4695	0.5459	0.5114	0.5459	0.5400	0.5218	0.7801
0.4415	2.4774	3800	0.4607	0.5639	0.5358	0.5639	0.5457	0.5335	0.7843
0.4335	2.6078	4000	0.4649	0.5525	0.5283	0.5525	0.5354	0.5349	0.7830
0.4335	2.7382	4200	0.4676	0.5457	0.5225	0.5457	0.5370	0.5348	0.7854
0.4335	2.8687	4400	0.4581	0.5606	0.5272	0.5606	0.5482	0.5250	0.7854
0.4347	2.9991	4600	0.4612	0.5650	0.5336	0.5650	0.5425	0.5341	0.7853
0.4347	3.1291	4800	0.4654	0.5580	0.5302	0.5580	0.5410	0.5358	0.7856
0.4048	3.2596	5000	0.4659	0.5706	0.5452	0.5706	0.5478	0.5463	0.7873
0.4048	3.3900	5200	0.4627	0.5692	0.5346	0.5692	0.5538	0.5311	0.7859
0.4048	3.5204	5400	0.4733	0.5557	0.5371	0.5557	0.5354	0.5451	0.7858
0.3995	3.6509	5600	0.4755	0.5538	0.5267	0.5538	0.5426	0.5308	0.7857
0.3995	3.7813	5800	0.4759	0.5467	0.5238	0.5467	0.5383	0.5342	0.7860
0.4016	3.9117	6000	0.4698	0.5566	0.5302	0.5566	0.5392	0.5368	0.7859
0.4016	4.0417	6200	0.4786	0.5646	0.5389	0.5646	0.5463	0.5369	0.7830
0.4016	4.1722	6400	0.4840	0.5636	0.5342	0.5636	0.5409	0.5319	0.7814
0.3723	4.3026	6600	0.4760	0.5653	0.5431	0.5653	0.5435	0.5457	0.7855
0.3723	4.4330	6800	0.4821	0.5632	0.5340	0.5632	0.5460	0.5348	0.7829
0.3682	4.5635	7000	0.4882	0.5566	0.5333	0.5566	0.5431	0.5389	0.7826

Framework versions

• Transformers 4.53.1
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.2