Mortezaem/Sentiment_Analysis_PEFT_on_DistilBert

results

This model is a fine-tuned version of distilbert-base-uncased on the sentiment140 (https://huggingface.co/datasets/sentiment140) dataset. Due to time and computational limits, I used only 10% of the dataset with a data augmentation. Despite minimal training, model's accuracy improved by 59.5% in comparison with non-fine tuned DistilBert LLM!

It achieves the following results on the evaluation set:

• Loss: 0.4732
• Accuracy: 0.7753

Classification Scores:

• Fine Tuned DistilBert LLM with LoRa: Accuracy = 0.78, Precision = 0.77, Recall = 0.78, F1 Score = 0.78

• Non-Fine Tuned DistilBert LLM: Accuracy = 0.49, Precision = 0.45, Recall = 0.14, F1 Score = 0.22

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-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.5092	1.0	5688	0.4979	0.7591
0.4928	2.0	11376	0.4876	0.7662
0.4915	3.0	17064	0.4821	0.7711
0.4787	4.0	22752	0.4779	0.7731
0.4757	5.0	28440	0.4767	0.7746
0.473	6.0	34128	0.4743	0.775
0.4649	7.0	39816	0.4741	0.7751
0.4709	8.0	45504	0.4732	0.7753

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2