bce54cd90e01cac2364a339370166db7

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

• Loss: 4.0145
• Data Size: 1.0
• Epoch Runtime: 567.1341
• Accuracy: 0.8992
• F1 Macro: 0.8525

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	17.1774	0	7.0184	0.1104	0.0930
No log	1	500	14.7993	0.0078	10.6877	0.1124	0.0341
No log	2	1000	9.5851	0.0156	23.7214	0.2908	0.0751
No log	3	1500	7.4400	0.0312	41.3457	0.3458	0.0857
No log	4	2000	7.1245	0.0625	72.7670	0.2908	0.0751
0.5056	5	2500	7.4436	0.125	115.5125	0.2908	0.0751
6.6172	6	3000	6.3971	0.25	186.0425	0.3488	0.0862
0.7021	7	3500	3.5013	0.5	292.9728	0.7606	0.5990
1.4406	8.0	4000	1.5970	1.0	587.5051	0.8921	0.8342
0.8411	9.0	4500	1.3535	1.0	592.3738	0.8952	0.8470
0.789	10.0	5000	1.1432	1.0	617.2037	0.9068	0.8600
0.6935	11.0	5500	3.0502	1.0	643.2457	0.8841	0.8308
0.4444	12.0	6000	2.1737	1.0	642.5709	0.8992	0.8432
0.5002	13.0	6500	2.9713	1.0	552.1847	0.9022	0.8553
0.2812	14.0	7000	4.0145	1.0	567.1341	0.8992	0.8525

Framework versions

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