37777ce92a62984e92da2f0e4b3d88e9

This model is a fine-tuned version of albert/albert-base-v2 on the ccdv/patent-classification [abstract] dataset. It achieves the following results on the evaluation set:

• Loss: 1.4272
• Data Size: 1.0
• Epoch Runtime: 33.2288
• Accuracy: 0.6042
• F1 Macro: 0.5049

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	2.1316	0	2.5315	0.2147	0.0616
No log	1	781	2.0618	0.0078	2.9489	0.2067	0.0620
No log	2	1562	2.0185	0.0156	3.0082	0.2073	0.0385
No log	3	2343	2.0003	0.0312	3.4383	0.2071	0.0381
0.0451	4	3124	2.0061	0.0625	4.3506	0.1977	0.0655
1.8524	5	3905	1.7046	0.125	6.1783	0.3215	0.1178
1.6141	6	4686	1.5711	0.25	9.7297	0.3407	0.1926
1.3877	7	5467	1.3495	0.5	16.9945	0.5327	0.3351
1.2497	8.0	6248	1.3044	1.0	32.8743	0.5523	0.3795
1.1413	9.0	7029	1.2155	1.0	33.0805	0.6080	0.4462
1.0173	10.0	7810	1.2083	1.0	32.6860	0.6088	0.4699
0.9171	11.0	8591	1.2616	1.0	32.8181	0.6044	0.4844
0.8333	12.0	9372	1.3445	1.0	32.6416	0.6056	0.4892
0.6727	13.0	10153	1.3684	1.0	32.6106	0.6002	0.5285
0.6542	14.0	10934	1.4272	1.0	33.2288	0.6042	0.5049

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1