phi2-QA-Arabic-phi-darebah-arabicandeng-newtest

This model is a fine-tuned version of microsoft/phi-2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6664

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: 2.5e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 2500

Training results

Training Loss	Epoch	Step	Validation Loss
1.3593	0.44	100	1.2815
1.1837	0.89	200	1.1457
1.0762	1.33	300	1.0660
0.9915	1.78	400	1.0016
0.9238	2.22	500	0.9568
0.9002	2.67	600	0.9155
0.8271	3.11	700	0.8798
0.8767	3.56	800	0.8471
0.8311	4.0	900	0.8225
0.7959	4.44	1000	0.8006
0.7031	4.89	1100	0.7836
0.761	5.33	1200	0.7618
0.7417	5.78	1300	0.7498
0.7258	6.22	1400	0.7319
0.6924	6.67	1500	0.7187
0.6622	7.11	1600	0.7127
0.6583	7.56	1700	0.6993
0.6851	8.0	1800	0.6924
0.6771	8.44	1900	0.6868
0.6832	8.89	2000	0.6805
0.6856	9.33	2100	0.6755
0.6747	9.78	2200	0.6724
0.659	10.22	2300	0.6691
0.657	10.67	2400	0.6674
0.6745	11.11	2500	0.6664

Framework versions

• PEFT 0.7.2.dev0
• Transformers 4.38.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1