visobert_massive_v4

This model is a fine-tuned version of uitnlp/visobert on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.7084
• Slot P: 0.6573
• Slot R: 0.7443
• Slot F1: 0.6981
• Slot Exact Match: 0.6783
• Intent Acc: 0.8603

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: 5e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 256
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Slot P	Slot R	Slot F1	Slot Exact Match	Intent Acc
No log	1.0	45	16.5324	0.3429	0.1194	0.1771	0.3512	0.0654
72.9207	2.0	90	6.7376	0.5847	0.5786	0.5816	0.5721	0.5755
19.858	3.0	135	4.4954	0.6110	0.7040	0.6542	0.6390	0.7801
8.9732	4.0	180	3.9970	0.6564	0.7308	0.6916	0.6714	0.8264
5.7389	5.0	225	3.9234	0.6355	0.7383	0.6831	0.6719	0.8357
4.1383	6.0	270	3.9908	0.6735	0.7398	0.7051	0.6867	0.8549
3.0123	7.0	315	4.2218	0.6457	0.7299	0.6852	0.6758	0.8608
2.2138	8.0	360	4.4620	0.6467	0.7313	0.6864	0.6778	0.8598
1.6407	9.0	405	4.7084	0.6573	0.7443	0.6981	0.6783	0.8603

Framework versions

• Transformers 4.55.0
• Pytorch 2.7.0+cu126
• Datasets 3.6.0
• Tokenizers 0.21.4