Danieljava/whisper-small-finetuned-asr

1---2library_name: transformers3language:4- en5- yo6- fr7- ig8- ha9license: apache-2.010base_model: openai/whisper-small11tags:12- generated_from_trainer13metrics:14- wer15model-index:16- name: Whisper Small Multilingual - EN/YO/FR/IG/HA/ES17  results: []18---19 20<!-- This model card has been generated automatically according to the information the Trainer had access to. You21should probably proofread and complete it, then remove this comment. -->22 23# Whisper Small Multilingual - EN/YO/FR/IG/HA/ES24 25This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset.26It achieves the following results on the evaluation set:27- Loss: 0.765328- Wer: 0.421329 30## Model description31 32More information needed33 34## Intended uses & limitations35 36More information needed37 38## Training and evaluation data39 40More information needed41 42## Training procedure43 44### Training hyperparameters45 46The following hyperparameters were used during training:47- learning_rate: 1e-0548- train_batch_size: 1649- eval_batch_size: 1650- seed: 4251- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments52- lr_scheduler_type: constant_with_warmup53- lr_scheduler_warmup_steps: 5054- training_steps: 100055- mixed_precision_training: Native AMP56 57### Training results58 59| Training Loss | Epoch   | Step | Validation Loss | Wer    |60|:-------------:|:-------:|:----:|:---------------:|:------:|61| 0.0045        | 17.2414 | 500  | 0.6874          | 0.4526 |62| 0.0005        | 34.4828 | 1000 | 0.7653          | 0.4213 |63 64 65### Framework versions66 67- Transformers 5.0.068- Pytorch 2.10.0+cu12869- Datasets 4.0.070- Tokenizers 0.22.271