Persian-Text-Sentiment-Bert-LORA

This model is a Adapter for HooshvareLab/bert-base-parsbert-uncased on SeyedAli/Persian-Text-Sentiment dataset in Persian Sentment Analysis Task. It achieves the following results on the evaluation set:

Loss: 0.3427
Precision: 0.8579
Recall: 0.8543
F1-score: 0.8540
Accuracy: 0.8543

Model description

More information needed

Intended uses & limitations

This is how to use this model in an example

from peft import PeftModel
from transformers import pipeline
modelname="SeyedAli/Persian-Text-Sentiment-Bert-LORA"
tokenizer=AutoTokenizer.from_pretrained("HooshvareLab/bert-base-parsbert-uncased")
model=AutoModelForSequenceClassification.from_pretrained("HooshvareLab/bert-base-parsbert-uncased")
model = PeftModel.from_pretrained(model, modelname)
pipe = pipeline("text-classification", model=model,tokenizer=tokenizer)
pipe('خیلی کتاب خوبی بود')

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: 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	Precision	Recall	F1-score	Accuracy
0.3939	1.0	3491	0.3835	0.8457	0.8404	0.8398	0.8404
0.3722	2.0	6982	0.3677	0.8513	0.8457	0.8451	0.8457
0.3553	3.0	10473	0.3576	0.8539	0.8495	0.8491	0.8495
0.3618	4.0	13964	0.3525	0.8546	0.8513	0.8509	0.8513
0.3534	5.0	17455	0.3485	0.8557	0.8521	0.8517	0.8521
0.3423	6.0	20946	0.3470	0.8562	0.8530	0.8526	0.8530
0.3455	7.0	24437	0.3453	0.8573	0.8535	0.8531	0.8535
0.347	8.0	27928	0.3428	0.8575	0.8539	0.8535	0.8539
0.344	9.0	31419	0.3429	0.8578	0.8546	0.8542	0.8546
0.335	10.0	34910	0.3427	0.8579	0.8543	0.8540	0.8543

Framework versions

Transformers 4.35.1
Pytorch 2.1.0+cu118
Datasets 2.14.6
Tokenizers 0.14.1