Hugging Face's logo

fydhfzh
/
ast-arabic

TensorBoard
Safetensors
audio-spectrogram-transformer
Generated from Trainer
Model card Files Metrics Community
ast-arabic / README.md
fydhfzh's picture
fydhfzh
End of training
4c15971 verified
preview code
|
raw
history blame
5.59 kB
metadata
license: bsd-3-clause
base_model: MIT/ast-finetuned-speech-commands-v2
tags:
  - generated_from_trainer
metrics:
  - accuracy
  - precision
  - recall
  - f1
model-index:
  - name: ast-arabic
    results: []

ast-arabic

This model is a fine-tuned version of MIT/ast-finetuned-speech-commands-v2 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5734
  • Accuracy: 0.8518
  • Precision: 0.8598
  • Recall: 0.8518
  • F1: 0.8430
  • Binary: 0.8957

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: 0.0001
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Accuracy Precision Recall F1 Binary
No log 0.19 50 4.3695 0.0162 0.0086 0.0162 0.0095 0.1469
No log 0.38 100 3.3975 0.2480 0.2204 0.2480 0.2000 0.4747
No log 0.58 150 1.9473 0.5067 0.4987 0.5067 0.4568 0.6574
No log 0.77 200 1.2409 0.6685 0.6872 0.6685 0.6429 0.7704
No log 0.96 250 0.9154 0.7493 0.7869 0.7493 0.7383 0.8267
No log 1.15 300 0.7568 0.7601 0.7885 0.7601 0.7487 0.8342
No log 1.34 350 0.6875 0.7951 0.8316 0.7951 0.7902 0.8558
No log 1.53 400 0.6592 0.8032 0.8336 0.8032 0.7964 0.8633
No log 1.73 450 0.6041 0.8059 0.8317 0.8059 0.7943 0.8663
1.8008 1.92 500 0.6096 0.8194 0.8487 0.8194 0.8087 0.8747
1.8008 2.11 550 0.5057 0.8410 0.8677 0.8410 0.8395 0.8887
1.8008 2.3 600 0.5269 0.8248 0.8642 0.8248 0.8234 0.8757
1.8008 2.49 650 0.4551 0.8598 0.8851 0.8598 0.8598 0.9022
1.8008 2.68 700 0.4962 0.8571 0.8766 0.8571 0.8552 0.9022
1.8008 2.88 750 0.5232 0.8464 0.8737 0.8464 0.8463 0.8935
1.8008 3.07 800 0.5889 0.8248 0.8583 0.8248 0.8189 0.8774
1.8008 3.26 850 0.5301 0.8571 0.8811 0.8571 0.8547 0.8992
1.8008 3.45 900 0.4731 0.8625 0.8812 0.8625 0.8614 0.9049
1.8008 3.64 950 0.4917 0.8544 0.8782 0.8544 0.8522 0.8992
0.2592 3.84 1000 0.4909 0.8625 0.8895 0.8625 0.8629 0.9049
0.2592 4.03 1050 0.5065 0.8544 0.8790 0.8544 0.8545 0.8992
0.2592 4.22 1100 0.5261 0.8437 0.8790 0.8437 0.8435 0.8906
0.2592 4.41 1150 0.5140 0.8598 0.8935 0.8598 0.8591 0.9019
0.2592 4.6 1200 0.4787 0.8760 0.8979 0.8760 0.8751 0.9132
0.2592 4.79 1250 0.5090 0.8652 0.8861 0.8652 0.8637 0.9049
0.2592 4.99 1300 0.5697 0.8437 0.8753 0.8437 0.8416 0.8898
0.2592 5.18 1350 0.5416 0.8625 0.8868 0.8625 0.8618 0.9038
0.2592 5.37 1400 0.5563 0.8518 0.8767 0.8518 0.8513 0.8962
0.2592 5.56 1450 0.5191 0.8679 0.8835 0.8679 0.8655 0.9097
0.0549 5.75 1500 0.4890 0.8787 0.9009 0.8787 0.8768 0.9164
0.0549 5.94 1550 0.4928 0.8814 0.9090 0.8814 0.8820 0.9181
0.0549 6.14 1600 0.5669 0.8679 0.8912 0.8679 0.8685 0.9097
0.0549 6.33 1650 0.5215 0.8760 0.9003 0.8760 0.8758 0.9132
0.0549 6.52 1700 0.5385 0.8760 0.9052 0.8760 0.8764 0.9135
0.0549 6.71 1750 0.5916 0.8625 0.8908 0.8625 0.8603 0.9038
0.0549 6.9 1800 0.5863 0.8598 0.8798 0.8598 0.8591 0.9030
0.0549 7.09 1850 0.5955 0.8571 0.8844 0.8571 0.8574 0.9011
0.0549 7.29 1900 0.6003 0.8571 0.8867 0.8571 0.8549 0.9011

Framework versions

  • Transformers 4.38.2
  • Pytorch 2.3.0
  • Datasets 2.19.1
  • Tokenizers 0.15.1