scenario-kd-from-scratch-gold-silver-data-tweet_eval-sentiment-model-xlm-roberta

This model is a fine-tuned version of xlm-roberta-base on the tweet_eval dataset. It achieves the following results on the evaluation set:

• Loss: 3.0720
• Accuracy: 0.6755
• F1: 0.6420

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: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 6969

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
4.3095	0.7	1000	3.4542	0.638	0.5615
3.1112	1.4	2000	3.0357	0.682	0.6484
2.8791	2.1	3000	2.9641	0.677	0.6373
2.5821	2.81	4000	3.0110	0.6695	0.6303
2.1895	3.51	5000	3.0888	0.672	0.6542
1.9976	4.21	6000	3.1685	0.677	0.6447
1.918	4.91	7000	3.1264	0.6715	0.6543
1.6104	5.61	8000	2.9487	0.6715	0.6396
1.3811	6.31	9000	3.5539	0.6555	0.6384
1.4193	7.01	10000	3.1164	0.661	0.6316
1.2829	7.71	11000	3.0770	0.681	0.6506
1.132	8.42	12000	3.0204	0.672	0.6428
1.1429	9.12	13000	2.9914	0.6635	0.6289
1.0642	9.82	14000	3.0322	0.668	0.6422
0.9858	10.52	15000	3.0235	0.664	0.6180
0.934	11.22	16000	2.9512	0.68	0.6393
0.9485	11.92	17000	3.0720	0.6755	0.6420

Framework versions

• Transformers 4.33.3
• Pytorch 2.0.1
• Datasets 2.14.5
• Tokenizers 0.13.3