StanfordAIMI/GREEN-Phi2

Phi-2

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: 0.0781

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 32
• total_train_batch_size: 2048
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 12.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.2903	0.64	25	0.1770
0.1566	1.28	50	0.1319
0.1379	1.92	75	0.1253
0.1246	2.56	100	0.1165
0.1159	3.2	125	0.1049
0.1048	3.84	150	0.0998
0.0947	4.48	175	0.0949
0.0872	5.12	200	0.0906
0.0836	5.76	225	0.0890
0.0774	6.39	250	0.0850
0.0717	7.03	275	0.0827
0.0639	7.67	300	0.0807
0.0596	8.31	325	0.0789
0.0555	8.95	350	0.0773
0.0498	9.59	375	0.0777
0.0491	10.23	400	0.0781
0.0467	10.87	425	0.0780
0.0459	11.51	450	0.0781

Framework versions

• Transformers 4.38.1
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2