kugler/gbert-large-AmDi-synset-classifier

results

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

• Loss: 0.5650
• Accuracy: 0.8403
• F1: 0.8328
• Precision: 0.8416
• Recall: 0.8403
• F1 Macro: 0.6886
• Precision Macro: 0.6871
• Recall Macro: 0.7119
• F1 Micro: 0.8403
• Precision Micro: 0.8403
• Recall Micro: 0.8403

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: 20
• eval_batch_size: 20
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 80
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall	F1 Macro	Precision Macro	Recall Macro	F1 Micro	Precision Micro	Recall Micro
3.1687	0.3891	100	1.7221	0.6317	0.5563	0.5358	0.6317	0.2779	0.2937	0.2938	0.6317	0.6317	0.6317
1.2239	0.7782	200	0.8836	0.7856	0.7633	0.7696	0.7856	0.5175	0.5077	0.5567	0.7856	0.7856	0.7856
0.7758	1.1673	300	0.7089	0.8107	0.7922	0.7939	0.8107	0.5917	0.5889	0.6185	0.8107	0.8107	0.8107
0.6436	1.5564	400	0.6498	0.8250	0.8136	0.8220	0.8250	0.6330	0.6331	0.6563	0.8250	0.8250	0.8250
0.5815	1.9455	500	0.6037	0.8300	0.8227	0.8338	0.8300	0.6583	0.6478	0.6890	0.8300	0.8300	0.8300
0.4695	2.3346	600	0.5771	0.8389	0.8319	0.8409	0.8389	0.6729	0.6688	0.6984	0.8389	0.8389	0.8389
0.4336	2.7237	700	0.5724	0.8362	0.8280	0.8395	0.8362	0.6753	0.6682	0.7038	0.8362	0.8362	0.8362
0.4135	3.1128	800	0.5650	0.8403	0.8328	0.8416	0.8403	0.6886	0.6871	0.7119	0.8403	0.8403	0.8403

Framework versions

• Transformers 4.45.2
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.20.3