CoolFace
Modelpublic

pere/whisper-NST2-unfreeze-constanti-low-lr

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face
0likes12downloads
Model Card

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

whisper-NST2-unfreeze-constanti-low-lr

This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.3562
  • —Wer: 8.5519

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: 1e-05
  • —trainbatchsize: 96
  • —evalbatchsize: 16
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —lrschedulerwarmup_steps: 1000
  • —training_steps: 20000
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossWer
0.19010.0510000.306914.8233
0.13230.120000.268711.2885
0.11370.1530000.262010.8324
0.10220.240000.29769.0080
0.09370.2550000.25849.5781
0.08750.360000.270420.2965
0.05921.0570000.27519.0080
0.04881.180000.27788.6659
0.04751.1590000.27929.4641
0.04391.2100000.28808.3238
0.04251.25110000.29548.5519
0.04161.3120000.289620.2965
0.02892.05130000.29907.9818
0.02292.1140000.30277.4116
0.02482.15150000.29688.6659
0.02252.2160000.31008.5519
0.02222.25170000.31329.3501
0.02192.3180000.32307.6397
0.01623.04190000.33809.8062
0.01323.09200000.35628.5519

Framework versions

  • —Transformers 4.25.0.dev0
  • —Pytorch 1.13.0+cu117
  • —Datasets 2.6.1
  • —Tokenizers 0.13.1