dc5d6d2d5ee2a2b31e2723fbc7582c3f

This model is a fine-tuned version of distilbert/distilgpt2 on the ccdv/patent-classification [abstract] dataset. It achieves the following results on the evaluation set:

• Loss: 1.2837
• Data Size: 1.0
• Epoch Runtime: 34.8389
• Accuracy: 0.6422
• F1 Macro: 0.5852

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	4.9206	0	3.3957	0.0503	0.0224
No log	1	781	2.1151	0.0078	4.0344	0.2029	0.1033
No log	2	1562	1.9476	0.0156	3.9425	0.2700	0.1113
No log	3	2343	1.6904	0.0312	4.4278	0.3894	0.2449
0.0464	4	3124	1.3884	0.0625	5.4914	0.5076	0.3563
1.3424	5	3905	1.2855	0.125	7.4379	0.5517	0.4345
1.2254	6	4686	1.1242	0.25	11.2880	0.6090	0.5264
1.0847	7	5467	1.0789	0.5	19.0538	0.6234	0.5611
0.9485	8.0	6248	1.0052	1.0	34.8075	0.6500	0.5816
0.8411	9.0	7029	1.0070	1.0	34.9952	0.6520	0.5848
0.7349	10.0	7810	1.0669	1.0	34.0879	0.6466	0.5885
0.6035	11.0	8591	1.1455	1.0	34.3573	0.6416	0.5821
0.4796	12.0	9372	1.2837	1.0	34.8389	0.6422	0.5852

Framework versions

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