postbot/bert_uncased_tiny_2xthicc-multi-emails-hq

bert_uncased_L-4_H-128_A-2-mlm-multi-emails-hq

This model is a fine-tuned version of google/bert_uncased_L-4_H-128_A-2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.8524
• Accuracy: 0.5077

Model description

Double the layers of BERT-tiny, fine-tuned on email data for eight epochs.

Intended uses & limitations

• This is primarily an example/test

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0003
• train_batch_size: 8
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 8.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.5477	0.99	141	3.2637	0.4551
3.3307	1.99	282	3.0873	0.4785
3.252	2.99	423	2.9842	0.4911
3.1415	3.99	564	2.9230	0.4995
3.0903	4.99	705	2.8625	0.5070
3.0996	5.99	846	2.8615	0.5087
3.0641	6.99	987	2.8407	0.5120
3.0514	7.99	1128	2.8524	0.5077

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 2.0.0.dev20230129+cu118
• Datasets 2.8.0
• Tokenizers 0.13.1