phi-2-heart

This model is a fine-tuned version of microsoft/phi-2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6017

• Accuracy: 0.6667

• Report: precision recall f1-score support

  absence 0.63 0.98 0.77 45 presence 0.91 0.28 0.43 36

  accuracy 0.67 81 macro avg 0.77 0.63 0.60 81

weighted avg 0.75 0.67 0.61 81

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: 2e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Report
No log	1.0	48	0.7215	0.5556	precision recall f1-score support

 absence       0.56      1.00      0.71        45
presence       0.00      0.00      0.00        36

accuracy                           0.56        81

macro avg 0.28 0.50 0.36 81 weighted avg 0.31 0.56 0.40 81 | | No log | 2.0 | 96 | 0.6567 | 0.5802 | precision recall f1-score support

 absence       0.57      1.00      0.73        45
presence       1.00      0.06      0.11        36

accuracy                           0.58        81

macro avg 0.78 0.53 0.42 81 weighted avg 0.76 0.58 0.45 81 | | No log | 3.0 | 144 | 0.6312 | 0.6667 | precision recall f1-score support

 absence       0.63      0.96      0.76        45
presence       0.85      0.31      0.45        36

accuracy                           0.67        81

macro avg 0.74 0.63 0.61 81 weighted avg 0.73 0.67 0.62 81 | | No log | 4.0 | 192 | 0.6253 | 0.5926 | precision recall f1-score support

 absence       0.58      1.00      0.73        45
presence       1.00      0.08      0.15        36

accuracy                           0.59        81

macro avg 0.79 0.54 0.44 81 weighted avg 0.76 0.59 0.47 81 | | No log | 5.0 | 240 | 0.6017 | 0.6667 | precision recall f1-score support

 absence       0.63      0.98      0.77        45
presence       0.91      0.28      0.43        36

accuracy                           0.67        81

macro avg 0.77 0.63 0.60 81 weighted avg 0.75 0.67 0.61 81 |

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.1