CodingQueen13/whisper-tiny-en

1---2library_name: transformers3license: apache-2.04base_model: openai/whisper-tiny5tags:6- generated_from_trainer7datasets:8- PolyAI/minds149metrics:10- wer11model-index:12- name: whisper-tiny-en13  results:14  - task:15      name: Automatic Speech Recognition16      type: automatic-speech-recognition17    dataset:18      name: PolyAI/minds1419      type: PolyAI/minds1420      config: en-US21      split: train[450:]22      args: en-US23    metrics:24    - name: Wer25      type: wer26      value: 0.328217237308146427---28 29<!-- This model card has been generated automatically according to the information the Trainer had access to. You30should probably proofread and complete it, then remove this comment. -->31 32# whisper-tiny-en33 34This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the PolyAI/minds14 dataset.35It achieves the following results on the evaluation set:36- Loss: 0.662637- Wer Ortho: 0.327038- Wer: 0.328239 40## Model description41 42More information needed43 44## Intended uses & limitations45 46More information needed47 48## Training and evaluation data49 50More information needed51 52## Training procedure53 54### Training hyperparameters55 56The following hyperparameters were used during training:57- learning_rate: 1e-0558- train_batch_size: 1659- eval_batch_size: 1660- seed: 4261- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-0862- lr_scheduler_type: constant_with_warmup63- lr_scheduler_warmup_steps: 5064- training_steps: 50065- mixed_precision_training: Native AMP66 67### Training results68 69| Training Loss | Epoch   | Step | Validation Loss | Wer Ortho | Wer    |70|:-------------:|:-------:|:----:|:---------------:|:---------:|:------:|71| 0.0006        | 17.8571 | 500  | 0.6626          | 0.3270    | 0.3282 |72 73 74### Framework versions75 76- Transformers 4.44.277- Pytorch 2.4.0+cu12178- Datasets 3.0.079- Tokenizers 0.19.180