61a9d4e5a3f5c2e482558bd336a0fca5

This model is a fine-tuned version of distilbert/distilgpt2 on the dair-ai/emotion [split] dataset. It achieves the following results on the evaluation set:

• Loss: 0.1776
• Data Size: 1.0
• Epoch Runtime: 21.2193
• Accuracy: 0.9264
• F1 Macro: 0.8772

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	5.7943	0	1.7046	0.1119	0.0335
No log	1	500	2.9745	0.0078	1.9954	0.2394	0.1062
No log	2	1000	1.8283	0.0156	2.2345	0.3347	0.1430
No log	3	1500	1.5442	0.0312	2.4292	0.3921	0.1494
No log	4	2000	1.3275	0.0625	3.0385	0.5101	0.2806
0.0797	5	2500	0.6951	0.125	4.2789	0.7470	0.5549
0.4781	6	3000	0.2765	0.25	6.6758	0.8931	0.8338
0.0358	7	3500	0.2785	0.5	11.4581	0.9133	0.8732
0.1806	8.0	4000	0.1357	1.0	21.8336	0.9284	0.8861
0.135	9.0	4500	0.1582	1.0	22.2750	0.9299	0.8889
0.1093	10.0	5000	0.1762	1.0	22.0260	0.9325	0.8951
0.084	11.0	5500	0.1902	1.0	22.2880	0.9309	0.8882
0.0877	12.0	6000	0.1776	1.0	21.2193	0.9264	0.8772

Framework versions

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