distilbert-base-uncased-finetuned-emotion-01

This model is a fine-tuned version of distilbert-base-uncased on the emotion dataset. It achieves the following results on the evaluation set:

• Loss: 0.2055
• Accuracy: 0.935
• F1: 0.9350

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: 2e-05
• train_batch_size: 768
• eval_batch_size: 768
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.0976	1.0	21	0.1546	0.9335	0.9334
0.0852	2.0	42	0.1595	0.936	0.9357
0.0747	3.0	63	0.1575	0.939	0.9394
0.0681	4.0	84	0.1648	0.938	0.9379
0.0596	5.0	105	0.1584	0.9335	0.9335
0.055	6.0	126	0.1800	0.934	0.9343
0.0518	7.0	147	0.1735	0.935	0.9353
0.0483	8.0	168	0.1847	0.936	0.9362
0.0469	9.0	189	0.1849	0.936	0.9359
0.0413	10.0	210	0.1871	0.935	0.9350
0.0414	11.0	231	0.1927	0.934	0.9337
0.0345	12.0	252	0.1969	0.9355	0.9354
0.0319	13.0	273	0.1942	0.9365	0.9366
0.0297	14.0	294	0.1979	0.937	0.9372
0.0283	15.0	315	0.1978	0.936	0.9359
0.0272	16.0	336	0.2012	0.9365	0.9364
0.0268	17.0	357	0.2039	0.9375	0.9374
0.0264	18.0	378	0.2059	0.937	0.9370
0.0234	19.0	399	0.2062	0.9365	0.9365
0.0232	20.0	420	0.2055	0.935	0.9350

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.1+cu118
• Datasets 2.20.0
• Tokenizers 0.19.1