README.md

metadata

language: ar
license: apache-2.0
library_name: peft
base_model: UBC-NLP/MARBERT
tags:
  - arabic
  - dialect-classification
  - lora

HammaLoRAMarBert

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	1.51721	1.50726	0.670392	0.685955	0.647908	0.695828	0.670392
2	0.827407	0.804686	0.779283	0.790449	0.779526	0.787574	0.779283
3	0.624589	0.617633	0.815747	0.823596	0.815815	0.818754	0.815747
4	0.577044	0.593563	0.822927	0.821348	0.824907	0.835161	0.822927
5	0.504094	0.535676	0.839036	0.834831	0.839583	0.842469	0.839036
6	0.46799	0.520281	0.849213	0.835955	0.850536	0.855303	0.849213
7	0.445317	0.510596	0.854708	0.840449	0.855552	0.858046	0.854708
8	0.428012	0.501261	0.858142	0.842135	0.859003	0.862191	0.858142
9	0.412287	0.491676	0.864635	0.848315	0.865268	0.868648	0.864635
10	0.400929	0.497091	0.868194	0.847753	0.8693	0.872337	0.868194
11	0.395328	0.506237	0.868319	0.840449	0.870433	0.87781	0.868319
12	0.378038	0.483877	0.874813	0.847191	0.875232	0.877259	0.874813
13	0.3727	0.488525	0.874313	0.841573	0.875207	0.878724	0.874313
14	0.366197	0.482607	0.878059	0.85	0.878635	0.880364	0.878059
15	0.365844	0.485294	0.878247	0.851124	0.879	0.88123	0.878247

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/HammaLoRAMarBert",
    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%})")