4f4bab2bb9a7aca2a69f38ef5018c3e0

This model is a fine-tuned version of Qwen/Qwen2.5-3B on the dair-ai/emotion [split] dataset. It achieves the following results on the evaluation set:

• Loss: 0.8603
• Data Size: 1.0
• Epoch Runtime: 156.4025
• Accuracy: 0.9249
• F1 Macro: 0.8725

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	24.6940	0	6.7166	0.1411	0.0546
No log	1	500	8.2605	0.0078	7.4637	0.3528	0.1034
No log	2	1000	5.9785	0.0156	13.3145	0.4068	0.1854
No log	3	1500	3.4628	0.0312	19.7901	0.7092	0.5740
No log	4	2000	3.0494	0.0625	28.2907	0.8180	0.7377
0.1952	5	2500	2.3765	0.125	43.1697	0.8463	0.8077
1.6105	6	3000	1.1382	0.25	51.0471	0.9098	0.8671
0.1917	7	3500	1.1078	0.5	85.2028	0.8987	0.8525
0.9004	8.0	4000	0.6006	1.0	161.0269	0.9244	0.8849
1.2301	9.0	4500	1.1154	1.0	154.7587	0.9113	0.8519
0.6026	10.0	5000	0.7333	1.0	154.7474	0.9315	0.8865
0.4862	11.0	5500	0.9760	1.0	156.2331	0.9254	0.8899
0.6039	12.0	6000	0.8603	1.0	156.4025	0.9249	0.8725

Framework versions

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