metadata

license: apache-2.0
base_model: openai/whisper-tiny
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: whisper-tiny-fr
    results: []

whisper-tiny-fr

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.8198
Wer: 0.8502

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
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 6250

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6223	1.0	250	0.7567	0.7225
0.475	2.0	500	0.6213	0.5461
0.2938	3.0	750	0.5860	0.5383
0.1613	4.0	1000	0.5903	0.4384
0.1026	5.0	1250	0.5992	0.4451
0.0615	6.0	1500	0.6322	0.5383
0.0422	7.0	1750	0.6398	0.4373
0.019	8.0	2000	0.6682	0.5239
0.0125	9.0	2250	0.6980	0.6681
0.0069	10.0	2500	0.7335	0.8679
0.0039	11.0	2750	0.7354	0.6238
0.0026	12.0	3000	0.7458	0.6315
0.0021	13.0	3250	0.7599	0.6715
0.0018	14.0	3500	0.7682	0.7103
0.0015	15.0	3750	0.7750	0.7081
0.0013	16.0	4000	0.7846	0.7125
0.0012	17.0	4250	0.7897	0.7114
0.001	18.0	4500	0.7962	0.9345
0.0009	19.0	4750	0.8001	0.7170
0.0009	20.0	5000	0.8074	0.8335
0.0008	21.0	5250	0.8107	0.8424
0.0007	22.0	5500	0.8152	0.8402
0.0007	23.0	5750	0.8181	0.8446
0.0007	24.0	6000	0.8187	0.8479
0.0007	25.0	6250	0.8198	0.8502

Framework versions

Transformers 4.32.0.dev0
Pytorch 2.0.1+cu118
Datasets 2.14.4
Tokenizers 0.13.3