59d699cfa46c42fb3e4bc3849a00588d

This model is a fine-tuned version of facebook/opt-1.3b on the ccdv/patent-classification [abstract] dataset. It achieves the following results on the evaluation set:

• Loss: 1.2516
• Data Size: 1.0
• Epoch Runtime: 145.1975
• Accuracy: 0.6120
• F1 Macro: 0.5637

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	2.5887	0	8.2413	0.1096	0.0575
No log	1	781	1.7558	0.0078	9.3337	0.3494	0.1793
No log	2	1562	1.6685	0.0156	12.0103	0.4161	0.2775
No log	3	2343	1.2349	0.0312	16.7655	0.5635	0.4343
0.0363	4	3124	1.1336	0.0625	22.9903	0.6056	0.5049
1.2368	5	3905	1.1255	0.125	33.2694	0.5935	0.5169
1.168	6	4686	1.0555	0.25	52.1426	0.6200	0.5504
1.0228	7	5467	1.1097	0.5	83.1411	0.6198	0.5463
0.9226	8.0	6248	1.1669	1.0	150.7202	0.6176	0.5519
0.8188	9.0	7029	1.0719	1.0	146.5441	0.6420	0.5931
0.6101	10.0	7810	1.2516	1.0	145.1975	0.6120	0.5637

Framework versions

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