mobilebert_sa_GLUE_Experiment_mnli_128

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

• Loss: 0.8825
• Accuracy: 0.5957

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: 128
• eval_batch_size: 128
• 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

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.0129	1.0	3068	0.9529	0.5438
0.9284	2.0	6136	0.9266	0.5593
0.8999	3.0	9204	0.9055	0.5775
0.8803	4.0	12272	0.8951	0.5854
0.8637	5.0	15340	0.8991	0.5886
0.8472	6.0	18408	0.8907	0.5913
0.8309	7.0	21476	0.8940	0.5928
0.814	8.0	24544	0.8880	0.5988
0.7988	9.0	27612	0.8776	0.6022
0.7825	10.0	30680	0.8958	0.6022
0.7662	11.0	33748	0.8835	0.6061
0.7504	12.0	36816	0.9004	0.6041
0.7359	13.0	39884	0.9252	0.6
0.7204	14.0	42952	0.9131	0.6007

Framework versions

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