metadata

license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
datasets:
  - dataset_whisper
metrics:
  - wer
model-index:
  - name: Transcriber-small
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: dataset_whisper
          type: dataset_whisper
          config: default
          split: test
          args: default
        metrics:
          - name: Wer
            type: wer
            value: 97.23577235772358

Transcriber-small

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

Loss: 3.0153
Wer: 97.2358

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: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 4000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.6006	4.02	100	2.6681	99.9350
1.6004	8.04	200	2.1138	107.2846
1.0072	12.06	300	1.9609	129.9187
0.5229	16.08	400	2.0901	119.0894
0.2155	20.1	500	2.2948	105.9187
0.0743	24.12	600	2.3731	100.6829
0.0292	28.14	700	2.5375	118.0813
0.0169	32.16	800	2.5601	108.0650
0.0121	36.18	900	2.6491	102.7642
0.008	40.2	1000	2.6436	94.3415
0.0046	44.22	1100	2.7131	89.8211
0.0021	48.24	1200	2.7516	96.9106
0.0012	52.26	1300	2.7878	95.3496
0.0009	56.28	1400	2.8137	97.6260
0.0008	60.3	1500	2.8333	94.2439
0.0007	64.32	1600	2.8514	90.1463
0.0006	68.34	1700	2.8667	95.3821
0.0006	72.36	1800	2.8813	98.0488
0.0005	76.38	1900	2.8932	98.8618
0.0005	80.4	2000	2.9056	98.9268
0.0004	84.42	2100	2.9156	96.7805
0.0004	88.44	2200	2.9251	96.7805
0.0004	92.46	2300	2.9343	97.8211
0.0003	96.48	2400	2.9439	97.8537
0.0003	100.5	2500	2.9516	97.1057
0.0003	104.52	2600	2.9597	98.1138
0.0003	108.54	2700	2.9671	96.4228
0.0003	112.56	2800	2.9733	99.1870
0.0003	116.58	2900	2.9791	102.2764
0.0003	120.6	3000	2.9860	101.2033
0.0002	124.62	3100	2.9903	98.9919
0.0002	128.64	3200	2.9953	98.3415
0.0002	132.66	3300	2.9996	99.8699
0.0002	136.68	3400	3.0034	100.1301
0.0002	140.7	3500	3.0070	98.7317
0.0002	144.72	3600	3.0093	97.1382
0.0002	148.74	3700	3.0118	98.3740
0.0002	152.76	3800	3.0136	96.8130
0.0002	156.78	3900	3.0153	96.8780
0.0002	160.8	4000	3.0153	97.2358

Framework versions

Transformers 4.32.0.dev0
Pytorch 1.12.1+cu113
Datasets 2.14.1
Tokenizers 0.13.3