Hugging Face's logo

Lakoc
/
DeCRED_linear_mixing_tuning

Transformers
Safetensors
joint_aed_ctc_speech-encoder-decoder
Generated from Trainer
Inference Endpoints
Model card Files Community
DeCRED_linear_mixing_tuning / README.md
Lakoc's picture
Lakoc
Upload README.md with huggingface_hub
33c380c verified
preview code
|
raw
history blame
3.58 kB
metadata
tags:
  - generated_from_trainer
base_model: Lakoc/DeCRED_small_cv_2
datasets:
  - common_voice_13_0
metrics:
  - wer
model-index:
  - name: DeCRED_linear_mixing_tuning
    results: []

DeCRED_linear_mixing_tuning

This model is a fine-tuned version of Lakoc/DeCRED_small_cv_2 on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

  • Loss: 1.0590
  • Cer: 0.0632
  • Wer: 0.1471
  • Mer: 0.1444
  • Wil: 0.2408
  • Wip: 0.7592
  • Hits: 23158
  • Substitutions: 2931
  • Deletions: 484
  • Insertions: 494

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: 0.05
  • train_batch_size: 256
  • eval_batch_size: 64
  • seed: 42
  • distributed_type: multi-GPU
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 1024
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 50.0

Training results

Training Loss Epoch Step Validation Loss Cer Wer Mer Wil Wip Hits Substitutions Deletions Insertions
1.185 2.67 20 1.1215 0.0646 0.1520 0.1492 0.2479 0.7521 23030 3013 530 496
1.0639 5.33 40 1.0717 0.0634 0.1480 0.1453 0.2420 0.7580 23124 2939 510 483
1.0957 8.0 60 1.0612 0.0628 0.1464 0.1438 0.2402 0.7598 23162 2929 482 480
1.0936 10.67 80 1.0595 0.0631 0.1469 0.1442 0.2407 0.7593 23158 2934 481 488
1.0804 13.33 100 1.0591 0.0630 0.1468 0.1441 0.2405 0.7595 23164 2929 480 492
1.1044 16.0 120 1.0591 0.0631 0.1469 0.1442 0.2405 0.7595 23162 2929 482 492
1.0836 18.67 140 1.0589 0.0631 0.1468 0.1442 0.2405 0.7595 23163 2929 481 492
1.0924 21.33 160 1.0590 0.0632 0.1471 0.1444 0.2408 0.7592 23158 2931 484 494
1.1048 24.0 180 1.0590 0.0632 0.1471 0.1444 0.2408 0.7592 23158 2931 484 494
1.0858 26.67 200 1.0589 0.0631 0.1470 0.1443 0.2407 0.7593 23160 2929 484 494
1.0953 29.33 220 1.0589 0.0631 0.1468 0.1442 0.2405 0.7595 23163 2929 481 492
1.1308 32.0 240 1.0590 0.0632 0.1471 0.1444 0.2408 0.7592 23158 2931 484 494

Framework versions

  • Transformers 4.40.0.dev0
  • Pytorch 2.2.0+rocm5.6
  • Datasets 2.18.0
  • Tokenizers 0.15.2

Wandb run

https://wandb.ai/butspeechfit/decred_commonvoice_en/runs/DeCRED_linear_mixing_tuning