metadata

language:
  - en
tags:
  - generated_from_trainer
base_model: openai/whisper-small
datasets:
  - QEC
metrics:
  - wer
model-index:
  - name: whisper-small-quartr
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: Quartr Earnings Calls
          type: QEC
          args: 'config: en, split: test'
        metrics:
          - type: wer
            value: 32.834424695977546
            name: Wer

whisper-small-quartr

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

Loss: 0.6429
Wer: 32.8344

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: 8.120528078446462e-06
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine_with_restarts
lr_scheduler_warmup_steps: 84
training_steps: 1500
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6771	0.32	100	0.6577	25.7437
0.6533	0.64	200	0.6025	34.6804
0.5793	0.96	300	0.5784	24.3530
0.3872	1.28	400	0.5856	32.9592
0.4447	1.61	500	0.5646	23.2429
0.414	1.93	600	0.5616	70.7016
0.2489	2.25	700	0.5816	35.2666
0.2863	2.57	800	0.5853	24.7583
0.2698	2.89	900	0.5844	32.7409
0.1646	3.21	1000	0.6182	27.7892
0.174	3.53	1100	0.6228	36.2021
0.2021	3.85	1200	0.6269	35.9775
0.2239	4.17	1300	0.6367	35.2666
0.1551	4.49	1400	0.6420	32.4478
0.1351	4.82	1500	0.6429	32.8344

Framework versions

Transformers 4.40.0.dev0
Pytorch 2.2.1+cu121
Datasets 2.18.1.dev0
Tokenizers 0.15.2