distilbert-base-uncased-IMDB_distilbert

This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.6232

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: 16

Training results

Training Loss	Epoch	Step	Validation Loss
3.1531	1.0	1250	2.9545
2.9251	2.0	2500	2.8577
2.7865	3.0	3750	2.8460
2.692	4.0	5000	2.7769
2.611	5.0	6250	2.8373
2.5341	6.0	7500	2.7105
2.4887	7.0	8750	2.6864
2.4292	8.0	10000	2.6600
2.3524	9.0	11250	2.6872
2.3217	10.0	12500	2.6527
2.2961	11.0	13750	2.6659
2.2553	12.0	15000	2.6513
2.2066	13.0	16250	2.6443
2.1912	14.0	17500	2.5912
2.1703	15.0	18750	2.6312
2.1715	16.0	20000	2.6232

Framework versions

• Transformers 4.21.1
• Pytorch 1.12.0+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1