ad99451f712d2b37c7c2dfef0e6ad63c

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Llama-8B on the ccdv/patent-classification [abstract] dataset. It achieves the following results on the evaluation set:

• Loss: 9.7487
• Data Size: 1.0
• Epoch Runtime: 2285.2438
• Accuracy: 0.5543
• F1 Macro: 0.5018

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	18.9011	0	30.3469	0.0813	0.0667
No log	1	781	8.8661	0.0078	46.1728	0.1470	0.0440
No log	2	1562	8.6804	0.0156	66.5819	0.2222	0.0413
No log	3	2343	7.9394	0.0312	108.9432	0.2422	0.0849
0.2457	4	3124	6.8274	0.0625	167.1900	0.3387	0.1542
6.8446	5	3905	6.1833	0.125	293.1628	0.3446	0.1482
5.7782	6	4686	5.5002	0.25	536.5399	0.5038	0.3454
4.9144	7	5467	5.3860	0.5	1016.0841	0.5603	0.4106
4.1792	8.0	6248	4.8623	1.0	1968.9650	0.5986	0.4794
3.3941	9.0	7029	5.2203	1.0	1949.2813	0.5994	0.4814
2.5111	10.0	7810	6.4367	1.0	1977.7298	0.5451	0.5161
1.8054	11.0	8591	6.8703	1.0	2169.3832	0.5801	0.5305
1.2608	12.0	9372	9.7487	1.0	2285.2438	0.5543	0.5018

Framework versions

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