fb63077eeae7e1358fff39bbfa9430e2

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B on the contemmcm/cls_20newsgroups dataset. It achieves the following results on the evaluation set:

• Loss: 2.8780
• Data Size: 1.0
• Epoch Runtime: 360.9313
• Accuracy: 0.8967
• F1 Macro: 0.8956

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	18.0267	0	25.2567	0.0517	0.0262
No log	1	499	13.9794	0.0078	28.6117	0.0920	0.0573
0.1592	2	998	8.2971	0.0156	31.7265	0.3972	0.3593
0.1915	3	1497	3.9569	0.0312	38.9826	0.6646	0.6542
0.1662	4	1996	3.5412	0.0625	50.4562	0.7427	0.7360
2.7036	5	2495	2.3552	0.125	73.3267	0.8180	0.8173
1.8031	6	2994	2.0861	0.25	116.4730	0.8458	0.8449
1.1871	7	3493	1.6320	0.5	196.9880	0.8669	0.8682
0.8305	8.0	3992	1.3696	1.0	358.6995	0.8876	0.8874
0.627	9.0	4491	1.3474	1.0	358.3273	0.9025	0.9034
0.4028	10.0	4990	1.7690	1.0	359.4797	0.9040	0.9016
0.3695	11.0	5489	2.1535	1.0	357.9018	0.8964	0.8937
0.3571	12.0	5988	2.1862	1.0	358.6151	0.8999	0.8991
0.3035	13.0	6487	2.8780	1.0	360.9313	0.8967	0.8956

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.1