metadata

license: apache-2.0
base_model: openai/whisper-tiny
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: torgo_tiny_finetune_M04_frozen_encoder
    results: []

torgo_tiny_finetune_M04_frozen_encoder

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: 0.2842
Wer: 39.5586

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: 16
eval_batch_size: 1
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7695	0.84	500	0.2502	52.2920
0.0895	1.69	1000	0.2592	39.9830
0.069	2.53	1500	0.2494	22.3260
0.0465	3.37	2000	0.2667	29.6265
0.0311	4.22	2500	0.2489	20.4584
0.0241	5.06	3000	0.2731	23.1749
0.0156	5.9	3500	0.2608	30.3056
0.0127	6.75	4000	0.2944	25.2971
0.0102	7.59	4500	0.2818	25.8913
0.008	8.43	5000	0.2610	25.1273
0.0079	9.27	5500	0.2632	24.6180
0.0054	10.12	6000	0.2776	29.4567
0.0047	10.96	6500	0.2758	28.0985
0.003	11.8	7000	0.2744	26.9949
0.0033	12.65	7500	0.2875	22.0713
0.0022	13.49	8000	0.2842	34.7199
0.0019	14.33	8500	0.2776	29.7963
0.0012	15.18	9000	0.2850	35.2292
0.0012	16.02	9500	0.2770	28.9474
0.0006	16.86	10000	0.2797	56.3667
0.0006	17.71	10500	0.2807	37.0119
0.0002	18.55	11000	0.2849	36.7572
0.0002	19.39	11500	0.2842	39.5586

Framework versions

Transformers 4.32.0
Pytorch 2.1.0+cu121
Datasets 2.14.7
Tokenizers 0.13.3