distilbert_add_GLUE_Experiment_mnli_192

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

• Loss: 0.9590
• Accuracy: 0.5259

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: 256
• eval_batch_size: 256
• seed: 10
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.096	1.0	1534	1.0536	0.4466
1.0362	2.0	3068	1.0527	0.4579
1.0213	3.0	4602	1.0341	0.4606
1.0085	4.0	6136	1.0170	0.4810
0.9971	5.0	7670	1.0013	0.4868
0.9877	6.0	9204	0.9913	0.4950
0.9805	7.0	10738	0.9872	0.4986
0.9726	8.0	12272	0.9822	0.5019
0.9658	9.0	13806	0.9812	0.5115
0.9566	10.0	15340	0.9761	0.5179
0.9439	11.0	16874	0.9650	0.5261
0.9336	12.0	18408	0.9616	0.5283
0.9232	13.0	19942	0.9620	0.5306
0.9145	14.0	21476	0.9663	0.5329
0.9056	15.0	23010	0.9491	0.5408
0.8986	16.0	24544	0.9567	0.5389
0.8903	17.0	26078	0.9488	0.5382
0.8835	18.0	27612	0.9739	0.5202
0.8769	19.0	29146	0.9761	0.5382
0.8701	20.0	30680	0.9504	0.5439
0.8633	21.0	32214	1.0074	0.5265
0.8576	22.0	33748	0.9874	0.5392

Framework versions

• Transformers 4.26.0
• Pytorch 1.14.0a0+410ce96
• Datasets 2.8.0
• Tokenizers 0.13.2