metadata

license: mit
base_model: microsoft/Phi-3-mini-4k-instruct
tags:
  - generated_from_trainer
model-index:
  - name: Phi0503B2
    results: []

Phi0503B2

This model is a fine-tuned version of microsoft/Phi-3-mini-4k-instruct on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0690

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: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 16
total_train_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine_with_restarts
lr_scheduler_warmup_steps: 100
num_epochs: 3
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
5.4837	0.09	10	5.4342
5.4537	0.18	20	5.2266
4.774	0.27	30	3.6419
2.4745	0.36	40	1.0488
0.5621	0.45	50	0.2015
0.1739	0.54	60	0.1465
0.1373	0.63	70	0.1350
0.1328	0.73	80	0.1258
0.1091	0.82	90	0.1152
0.1142	0.91	100	0.0968
0.0918	1.0	110	0.1021
0.0773	1.09	120	0.0807
0.0711	1.18	130	0.0793
0.0751	1.27	140	0.0661
0.06	1.36	150	0.0651
0.0647	1.45	160	0.0658
0.0577	1.54	170	0.0657
0.0575	1.63	180	0.0644
0.0534	1.72	190	0.0661
0.0594	1.81	200	0.0622
0.0473	1.9	210	0.0628
0.0522	1.99	220	0.0643
0.0402	2.08	230	0.0644
0.0436	2.18	240	0.0674
0.0343	2.27	250	0.0708
0.0358	2.36	260	0.0724
0.0411	2.45	270	0.0720
0.0359	2.54	280	0.0706
0.0366	2.63	290	0.0702
0.0397	2.72	300	0.0697
0.044	2.81	310	0.0692
0.0415	2.9	320	0.0688
0.037	2.99	330	0.0690

Framework versions

Transformers 4.36.0.dev0
Pytorch 2.1.2+cu121
Datasets 2.14.6
Tokenizers 0.14.1