gemma-2b-heart

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

gemma-2b-heart

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

• Loss: 0.4098

• Accuracy: 0.8222

• Report: precision recall f1-score support

  absence 0.82 0.86 0.84 98 presence 0.82 0.78 0.80 82

  accuracy 0.82 180 macro avg 0.82 0.82 0.82 180 weighted avg 0.82 0.82 0.82 180

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: 10

Training results

 absence       0.82      0.78      0.80        98
presence       0.75      0.79      0.77        82

accuracy                           0.78       180

macro avg 0.78 0.78 0.78 180 weighted avg 0.79 0.78 0.78 180 | | No log | 2.0 | 210 | 0.5009 | 0.7667 | precision recall f1-score support

 absence       0.87      0.67      0.76        98
presence       0.69      0.88      0.77        82

accuracy                           0.77       180

macro avg 0.78 0.78 0.77 180 weighted avg 0.79 0.77 0.77 180 | | No log | 3.0 | 315 | 0.4329 | 0.8 | precision recall f1-score support

 absence       0.85      0.77      0.81        98
presence       0.75      0.84      0.79        82

accuracy                           0.80       180

macro avg 0.80 0.80 0.80 180 weighted avg 0.81 0.80 0.80 180 | | No log | 4.0 | 420 | 0.4098 | 0.8222 | precision recall f1-score support

 absence       0.82      0.86      0.84        98
presence       0.82      0.78      0.80        82

accuracy                           0.82       180

macro avg 0.82 0.82 0.82 180 weighted avg 0.82 0.82 0.82 180 | | 0.553 | 5.0 | 525 | 0.5819 | 0.7778 | precision recall f1-score support

 absence       0.93      0.64      0.76        98
presence       0.69      0.94      0.79        82

accuracy                           0.78       180

macro avg 0.81 0.79 0.78 180 weighted avg 0.82 0.78 0.77 180 | | 0.553 | 6.0 | 630 | 0.5943 | 0.7722 | precision recall f1-score support

 absence       0.93      0.63      0.75        98
presence       0.68      0.94      0.79        82

accuracy                           0.77       180

macro avg 0.80 0.79 0.77 180 weighted avg 0.81 0.77 0.77 180 | | 0.553 | 7.0 | 735 | 0.5722 | 0.7611 | precision recall f1-score support

 absence       0.89      0.64      0.75        98
presence       0.68      0.90      0.77        82

accuracy                           0.76       180

macro avg 0.78 0.77 0.76 180 weighted avg 0.79 0.76 0.76 180 | | 0.553 | 8.0 | 840 | 0.6557 | 0.7667 | precision recall f1-score support

 absence       0.94      0.61      0.74        98
presence       0.67      0.95      0.79        82

accuracy                           0.77       180

macro avg 0.80 0.78 0.76 180 weighted avg 0.82 0.77 0.76 180 | | 0.553 | 9.0 | 945 | 0.6161 | 0.7667 | precision recall f1-score support

 absence       0.90      0.64      0.75        98
presence       0.68      0.91      0.78        82

accuracy                           0.77       180

macro avg 0.79 0.78 0.77 180 weighted avg 0.80 0.77 0.76 180 | | 0.4474 | 10.0 | 1050 | 0.6243 | 0.7778 | precision recall f1-score support

 absence       0.91      0.65      0.76        98
presence       0.69      0.93      0.79        82

accuracy                           0.78       180

macro avg 0.80 0.79 0.78 180 weighted avg 0.81 0.78 0.78 180 |

Framework versions

• PEFT 0.15.2
• Transformers 4.52.3
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.1