d_bert_v1

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

• Loss: 0.3069
• Accuracy: 0.8929
• F1: 0.8931
• Precision: 0.8942
• Recall: 0.8929

Model description

Discriminator model for semantically similar classes

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: 32
• eval_batch_size: 128
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.5
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
1.5322	0.16	500	1.1796	0.6822	0.6712	0.6833	0.6822
0.7762	0.32	1000	0.5988	0.8002	0.7985	0.7973	0.8002
0.5433	0.48	1500	0.4945	0.8290	0.8288	0.8295	0.8290
0.4879	0.64	2000	0.4819	0.8301	0.8319	0.8407	0.8301
0.4447	0.8	2500	0.4223	0.8496	0.8511	0.8542	0.8496
0.4187	0.96	3000	0.4062	0.8525	0.8541	0.8594	0.8525
0.3746	1.12	3500	0.3892	0.8657	0.8650	0.8654	0.8657
0.3615	1.28	4000	0.3829	0.8637	0.8656	0.8694	0.8637
0.3507	1.44	4500	0.3501	0.8735	0.8748	0.8784	0.8735
0.3369	1.6	5000	0.3900	0.8567	0.8601	0.8759	0.8567
0.332	1.76	5500	0.3247	0.8842	0.8850	0.8867	0.8842
0.3316	1.92	6000	0.3280	0.8807	0.8803	0.8816	0.8807
0.2858	2.08	6500	0.3257	0.8881	0.8879	0.8881	0.8881
0.2613	2.24	7000	0.3282	0.8850	0.8861	0.8889	0.8850
0.2575	2.4	7500	0.3209	0.8875	0.8881	0.8913	0.8875
0.241	2.56	8000	0.3204	0.8896	0.8905	0.8930	0.8896
0.2431	2.7200	8500	0.3225	0.8851	0.8862	0.8903	0.8851
0.2248	2.88	9000	0.3069	0.8929	0.8931	0.8942	0.8929

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0
• Datasets 3.0.0
• Tokenizers 0.19.1