metadata

language: ar
license: apache-2.0
library_name: peft
base_model: CAMeL-Lab/bert-base-arabic-camelbert-mix
tags:
  - arabic
  - dialect-classification
  - lora

HammaLoRACAMeLBert

Advanced Arabic Dialect Classification Model with Complete Training Metrics

Full Training History:

epoch	train_loss	eval_loss	train_accuracy	eval_accuracy	f1	precision	recall
1	1.04802	1.05081	0.637238	0.63764	0.6275	0.63901	0.637238
2	0.77121	0.781513	0.736139	0.735955	0.613924	0.630747	0.624063
3	0.620521	0.659047	0.788899	0.779775	0.791235	0.799805	0.788899
4	0.573822	0.644417	0.800762	0.776404	0.67889	0.688134	0.680569
5	0.49636	0.587736	0.824675	0.791573	0.826167	0.833785	0.824675
6	0.476073	0.589205	0.833292	0.792697	0.709791	0.724241	0.708417
7	0.429335	0.547112	0.849525	0.808427	0.850117	0.853328	0.849525
8	0.412983	0.544361	0.855519	0.805056	0.732394	0.747566	0.729645
9	0.388498	0.551354	0.863511	0.808989	0.863979	0.869072	0.863511
10	0.359349	0.52705	0.877185	0.81573	0.74493	0.757555	0.743194
11	0.34684	0.538555	0.879683	0.810674	0.880776	0.884118	0.879683
12	0.337791	0.538365	0.883304	0.807865	0.749058	0.760792	0.747378
13	0.328992	0.53789	0.886301	0.81236	0.887004	0.889841	0.886301
14	0.322374	0.536085	0.889423	0.81573	0.758437	0.767501	0.756868
15	0.320124	0.53761	0.890485	0.814607	0.891161	0.89363	0.890485

Label Mapping:

{0: 'Egypt', 1: 'Iraq', 2: 'Lebanon', 3: 'Morocco', 4: 'Saudi_Arabia', 5: 'Sudan', 6: 'Tunisia'}

USAGE Example:

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="Hamma-16/HammaLoRACAMeLBert",
    device="cuda" if torch.cuda.is_available() else "cpu"
)

sample_text = "شلونك اليوم؟"
result = classifier(sample_text)
print(f"Text: {sample_text}")
print(f"Predicted: {result[0]['label']} (confidence: {result[0]['score']:.1%})")