phi3.5-mini-adapter_v2

This model is a fine-tuned version of microsoft/Phi-3.5-mini-instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1344

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: 1e-05
• train_batch_size: 24
• eval_batch_size: 24
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
17.6422	0.4545	10	17.2166
13.0308	0.9091	20	12.5267
8.4662	1.3636	30	7.9483
2.4521	1.8182	40	1.3982
0.3917	2.2727	50	0.3427
0.3042	2.7273	60	0.3092
0.2051	3.1818	70	0.2451
0.2043	3.6364	80	0.2022
0.1599	4.0909	90	0.1907
0.1658	4.5455	100	0.1727
0.1527	5.0	110	0.1595
0.1281	5.4545	120	0.1501
0.1079	5.9091	130	0.1435
0.0896	6.3636	140	0.1369
0.097	6.8182	150	0.1340
0.0841	7.2727	160	0.1449
0.0771	7.7273	170	0.1344

Framework versions

• PEFT 0.11.1
• Transformers 4.43.1
• Pytorch 2.4.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1