saccade-v2-4m

This model is a fine-tuned version of aayushbist/saccade-tiny-100k on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1586
• Accuracy: 0.9389
• Keep F1: 0.9545
• Drop F1: 0.9072
• Macro F1: 0.9309
• Predicted Drop Rate: 0.3293
• Gold Drop Rate: 0.3288

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: 3e-05
• train_batch_size: 64
• eval_batch_size: 128
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Keep F1	Drop F1	Macro F1	Predicted Drop Rate	Gold Drop Rate
0.2646	0.3918	500	0.2424	0.9115	0.9362	0.8557	0.8959	0.2843	0.3288
0.2006	0.7837	1000	0.1701	0.9367	0.9531	0.9030	0.9280	0.3234	0.3288
0.1862	1.1755	1500	0.1626	0.9384	0.9542	0.9061	0.9302	0.3269	0.3288
0.1741	1.5674	2000	0.1593	0.9382	0.9538	0.9065	0.9302	0.3321	0.3288
0.1772	1.9592	2500	0.1586	0.9391	0.9546	0.9074	0.9310	0.3288	0.3288
0.1772	2.0	2552	0.1586	0.9389	0.9545	0.9072	0.9309	0.3293	0.3288

Framework versions

• Transformers 5.13.1
• Pytorch 2.11.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2