Sanrove/albert-spam-sms-classification-finetuned

training_dir

This model is a fine-tuned version of albert-base-v2 on an Spam Data Collection dataset. It achieves the following results on the evaluation set:

• Loss: 0.0393
• Accuracy: 0.9946
• F1 Score: 0.9946

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

SMS 1:

Message: Hey, I'll be there in 10 minutes. See you soon!

Label: label_0 (ham)

SMS 2:

Message: Congratulations! You've won a $1000 gift card. Claim it now by clicking the link.

Label: label_1 (spam)

In this SMS classification example, the first message is labeled as "label_0" because it appears to be a legitimate text message (ham) with someone informing they will arrive shortly. The second message is labeled as "label_1" because it is clearly spam, offering a prize and urging the recipient to click a link, which is a common characteristic of spam messages. The classification model uses these labels to identify and filter out spammy SMS messages, ensuring that legitimate messages reach the user's inbox (ham).

Training procedure

Colab

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Score
No log	1.0	244	0.1070	0.9785	0.9785
No log	2.0	488	0.0673	0.9880	0.9880
0.0885	3.0	732	0.0293	0.9946	0.9946
0.0885	4.0	976	0.0280	0.9964	0.9964
0.0306	5.0	1220	0.0355	0.9952	0.9952
0.0306	6.0	1464	0.0364	0.9952	0.9952
0.0087	7.0	1708	0.0448	0.9946	0.9946
0.0087	8.0	1952	0.0618	0.9922	0.9922
0.0047	9.0	2196	0.0420	0.9946	0.9946
0.0047	10.0	2440	0.0393	0.9946	0.9946

Framework versions

• Transformers 4.33.2
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3