adriansanz/te-zsc-synthetic

SYN_300524_epoch_5

This model is a fine-tuned version of projecte-aina/roberta-base-ca-v2-cased-te on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3372
• Accuracy: 0.98
• Precision: 0.9803
• Recall: 0.98
• F1: 0.9800
• Ratio: 0.488

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: 47
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.06
• lr_scheduler_warmup_steps: 4
• num_epochs: 1
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Ratio
0.3174	0.0533	10	0.3307	0.984	0.9840	0.984	0.9840	0.496
0.3202	0.1067	20	0.3258	0.986	0.9861	0.986	0.9860	0.494
0.3016	0.16	30	0.3282	0.986	0.9860	0.986	0.9860	0.504
0.3291	0.2133	40	0.3495	0.977	0.9774	0.977	0.9770	0.485
0.2942	0.2667	50	0.3602	0.973	0.9738	0.973	0.9730	0.479
0.3121	0.32	60	0.3586	0.973	0.9731	0.973	0.9730	0.493
0.3226	0.3733	70	0.3736	0.968	0.9681	0.968	0.9680	0.508
0.3226	0.4267	80	0.3515	0.979	0.9791	0.979	0.9790	0.493
0.3265	0.48	90	0.3697	0.97	0.9706	0.97	0.9700	0.482
0.3424	0.5333	100	0.3650	0.971	0.9717	0.971	0.9710	0.481
0.3348	0.5867	110	0.3502	0.976	0.9760	0.976	0.9760	0.496
0.3393	0.64	120	0.3441	0.978	0.9780	0.978	0.9780	0.496
0.3421	0.6933	130	0.3397	0.979	0.9791	0.979	0.9790	0.493
0.3319	0.7467	140	0.3412	0.979	0.9791	0.979	0.9790	0.493
0.3554	0.8	150	0.3416	0.977	0.9772	0.977	0.9770	0.489
0.3829	0.8533	160	0.3428	0.978	0.9785	0.978	0.9780	0.484
0.3631	0.9067	170	0.3396	0.979	0.9793	0.979	0.9790	0.487
0.3362	0.96	180	0.3376	0.98	0.9803	0.98	0.9800	0.488

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1