phi2-QA-Arabic-phi

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: 0.7778

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
1.1134	0.89	100	1.0092
0.8768	1.78	200	0.8800
0.7644	2.67	300	0.8329
0.7516	3.56	400	0.8081
0.6618	4.44	500	0.7909
0.6373	5.33	600	0.7845
0.6154	6.22	700	0.7688
0.6056	7.11	800	0.7716
0.5719	8.0	900	0.7662
0.5575	8.89	1000	0.7700
0.5302	9.78	1100	0.7689
0.5465	10.67	1200	0.7688
0.5321	11.56	1300	0.7719
0.5141	12.44	1400	0.7684
0.5033	13.33	1500	0.7716
0.4931	14.22	1600	0.7664
0.4882	15.11	1700	0.7739
0.4742	16.0	1800	0.7757
0.4701	16.89	1900	0.7717
0.4932	17.78	2000	0.7748
0.4665	18.67	2100	0.7734
0.4614	19.56	2200	0.7809
0.4669	20.44	2300	0.7793
0.4635	21.33	2400	0.7750
0.452	22.22	2500	0.7778

Framework versions

• PEFT 0.7.2.dev0
• Transformers 4.38.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0