bert-arxiv-classification – AI Model by kovalchuk026 | AlphaNeural AI

You can deploy this model and start earning money today!

kovalchuk026/
bert-arxiv-classification

like

0

text-classification

generated_from_trainer

distilbert/distilbert-base-uncased

autotrain_compatible

endpoints_compatible

Views

No views yet

bert-arxiv-classification

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: 0.5679
Accuracy: 0.808
F1 Micro: 0.808
F1 Macro: 0.7978
F1 Weighted: 0.8082

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: 32
eval_batch_size: 32
seed: 42
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: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Micro	F1 Macro	F1 Weighted
0.7623	1.0	1125	0.5694	0.8063	0.8063	0.7928	0.8032
0.6339	2.0	2250	0.5620	0.8067	0.8067	0.7977	0.8076
0.3244	3.0	3375	0.5520	0.818	0.818	0.8068	0.8171
0.3556	4.0	4500	0.5679	0.808	0.808	0.7978	0.8082

Framework versions

Transformers 4.51.0
Pytorch 2.6.0+cu124
Datasets 3.5.0
Tokenizers 0.21.1