gp-tar4/QA_FineTuned_ArabianGpt-01B

model_outputs

This model is a fine-tuned version of riotu-lab/ArabianGPT-01B on an arcd dataset(Arabic dataset). It achieves the following results on the evaluation set:

• Loss: 3.0808

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
3.8475	0.25	10	3.8987
4.0025	0.49	20	3.6754
3.6385	0.74	30	3.2931
3.2785	0.99	40	2.9471
2.1751	1.23	50	3.0013
1.8868	1.48	60	3.0324
1.9831	1.73	70	2.8470
1.8749	1.98	80	2.8488
1.0702	2.22	90	2.9369
0.8701	2.47	100	3.0490
0.8731	2.72	110	3.0774
0.8309	2.96	120	3.0808

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2