dalyaa/phi2-QA-darebah-6700

phi-2

This model is a fine-tuned version of microsoftl on the dalyaa/darebah6700 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8023

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
1.2748	0.15	100	1.1349
1.1095	0.29	200	1.0077
1.0138	0.44	300	0.9587
0.9506	0.58	400	0.9188
0.9047	0.73	500	0.8906
0.9017	0.87	600	0.8722
0.8872	1.02	700	0.8660
0.8744	1.16	800	0.8501
0.8221	1.31	900	0.8472
0.8356	1.45	1000	0.8380
0.8335	1.6	1100	0.8317
0.828	1.75	1200	0.8273
0.8307	1.89	1300	0.8231
0.8147	2.04	1400	0.8185
0.8012	2.18	1500	0.8139
0.7885	2.33	1600	0.8129
0.7831	2.47	1700	0.8102
0.7787	2.62	1800	0.8148
0.7921	2.76	1900	0.8083
0.7777	2.91	2000	0.8073
0.7766	3.05	2100	0.8045
0.7658	3.2	2200	0.8056
0.7651	3.35	2300	0.8044
0.7844	3.49	2400	0.8023
0.7876	3.64	2500	0.8023

Framework versions

• PEFT 0.8.2
• Transformers 4.38.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1