metadata

language:
  - he
license: apache-2.0
base_model: openai/whisper-tiny
tags:
  - hf-asr-leaderboard
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: he-cantillation
    results: []

he-cantillation

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

Loss: 0.1635
Wer: 15.8869
Avg Precision Exact: 0.8555
Avg Recall Exact: 0.8550
Avg F1 Exact: 0.8548
Avg Precision Letter Shift: 0.8805
Avg Recall Letter Shift: 0.8801
Avg F1 Letter Shift: 0.8799
Avg Precision Word Level: 0.8850
Avg Recall Word Level: 0.8845
Avg F1 Word Level: 0.8842
Avg Precision Word Shift: 0.9658
Avg Recall Word Shift: 0.9665
Avg F1 Word Shift: 0.9656
Precision Median Exact: 0.9333
Recall Median Exact: 0.9333
F1 Median Exact: 0.9375
Precision Max Exact: 1.0
Recall Max Exact: 1.0
F1 Max Exact: 1.0
Precision Min Exact: 0.0
Recall Min Exact: 0.0
F1 Min Exact: 0.0
Precision Min Letter Shift: 0.0
Recall Min Letter Shift: 0.0
F1 Min Letter Shift: 0.0
Precision Min Word Level: 0.0
Recall Min Word Level: 0.0
F1 Min Word Level: 0.0
Precision Min Word Shift: 0.1429
Recall Min Word Shift: 0.1111
F1 Min Word Shift: 0.125

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: 8
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 50
training_steps: 20000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Avg Precision Exact	Avg Recall Exact	Avg F1 Exact	Avg Precision Letter Shift	Avg Recall Letter Shift	Avg F1 Letter Shift	Avg Precision Word Level	Avg Recall Word Level	Avg F1 Word Level	Avg Precision Word Shift	Avg Recall Word Shift	Avg F1 Word Shift	Precision Median Exact	Recall Median Exact	F1 Median Exact	Precision Max Exact	Recall Max Exact	F1 Max Exact	Precision Min Word Shift	Recall Min Word Shift	F1 Min Word Shift
No log	8e-05	1	8.6830	101.5891	0.0009	0.0016	0.0010	0.0100	0.0077	0.0077	0.0054	0.0173	0.0076	0.0465	0.0380	0.0371	0.0	0.0	0.0	0.125	0.3333	0.1667	0.0	0.0	0.0
0.0958	0.32	4000	0.1924	25.9534	0.7565	0.7583	0.7566	0.7917	0.7936	0.7919	0.7983	0.8005	0.7986	0.9281	0.9324	0.9293	0.8571	0.8571	0.8571	1.0	1.0	1.0	0.0	0.0	0.0
0.0539	0.64	8000	0.1813	21.8625	0.7983	0.7994	0.7983	0.8279	0.8291	0.8279	0.8342	0.8359	0.8344	0.9437	0.9468	0.9445	0.9091	0.9091	0.9032	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0353	0.96	12000	0.1755	19.0909	0.8271	0.8284	0.8273	0.8558	0.8572	0.8560	0.8610	0.8624	0.8611	0.9537	0.9561	0.9542	0.9167	0.9167	0.9231	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0197	1.28	16000	0.1664	17.0288	0.8452	0.8444	0.8444	0.8711	0.8705	0.8703	0.8762	0.8756	0.8754	0.9609	0.9625	0.9610	0.9286	0.9286	0.9286	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0088	1.6	20000	0.1635	15.8869	0.8555	0.8550	0.8548	0.8805	0.8801	0.8799	0.8850	0.8845	0.8842	0.9658	0.9665	0.9656	0.9333	0.9333	0.9375	1.0	1.0	1.0	0.1429	0.1111	0.125

Framework versions

Transformers 4.41.2
Pytorch 2.2.1
Datasets 2.20.0
Tokenizers 0.19.1