output / README.md
Priyanship's picture
Training in progress, step 1000
8f736df verified
|
raw
history blame
3.98 kB
metadata
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: output
    results: []

Visualize in Weights & Biases

output

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 2.0007
  • Cer: 0.3962
  • Wer: 0.6069

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: 0.0006
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 300
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1500
  • training_steps: 3500
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Cer Wer
5.2909 0.16 100 5.7883 0.9933 1.0
4.4471 0.32 200 3.7597 0.9812 1.0
3.6763 0.48 300 3.7718 0.9769 1.0
3.1272 0.64 400 3.4647 0.9826 1.0
3.0699 0.8 500 2.7480 0.6002 0.8337
1.7389 0.96 600 2.4468 0.4717 0.6782
1.7013 1.12 700 2.1151 0.4489 0.6820
1.2995 1.28 800 2.0757 0.4160 0.6245
1.6852 1.44 900 1.9870 0.4112 0.6154
1.3997 1.6 1000 2.0007 0.3962 0.6069
1.768 1.76 1100 2.0712 0.4123 0.6448
2.5192 1.92 1200 2.5729 0.6884 0.9178
2.6077 2.08 1300 2.4078 0.4816 0.8066
2.6928 2.24 1400 2.3596 0.4904 0.7915
0.0 2.4 1500 2.4471 0.6019 0.8782
0.0 2.56 1600 2.4490 0.6112 0.8888
0.0 2.7200 1700 nan 1.0 1.0
0.0 2.88 1800 nan 1.0 1.0
0.0 3.04 1900 nan 1.0 1.0
0.0 3.2 2000 nan 1.0 1.0
0.0 3.36 2100 nan 1.0 1.0
0.0 3.52 2200 nan 1.0 1.0
0.0 3.68 2300 nan 1.0 1.0
0.0 3.84 2400 nan 1.0 1.0
0.0 4.0 2500 nan 1.0 1.0
0.0 4.16 2600 nan 1.0 1.0
0.0 4.32 2700 nan 1.0 1.0
0.0 4.48 2800 nan 1.0 1.0
0.0 4.64 2900 nan 1.0 1.0
0.0 4.8 3000 nan 1.0 1.0
0.0 4.96 3100 nan 1.0 1.0
0.0 5.12 3200 nan 1.0 1.0
0.0 5.28 3300 nan 1.0 1.0
0.0 5.44 3400 nan 1.0 1.0
0.0 5.6 3500 nan 1.0 1.0

Framework versions

  • Transformers 4.43.1
  • Pytorch 2.4.0
  • Datasets 2.20.0
  • Tokenizers 0.19.1