gemini_whisper_ipa

Fine-tuned openai/whisper-small for Santali IPA/Sanlish transliteration (speech-to-text).

Last updated: 2026-08-30T04:53:30.291684Z — after epoch 10.0

Dataset split (80/10/10)

  • train: 1547
  • validation: 193
  • test: 194

Latest evaluation metrics (epoch 10.0)

  • test_loss: 0.9323491454124451
  • test_wer: 34.66981132075472
  • test_cer: 8.268111579369858

Training hyperparameters

  • per_device_train_batch_size: 8
  • gradient_accumulation_steps: 2
  • effective_batch_size: 16
  • learning_rate: 3e-05
  • warmup_steps: 100
  • num_train_epochs: 10
  • gradient_checkpointing: True
  • fp16: True
  • eval_strategy: IntervalStrategy.EPOCH
  • save_strategy: SaveStrategy.EPOCH
  • save_total_limit: 2
  • load_best_model_at_end: True
  • metric_for_best_model: wer
  • predict_with_generate: True
  • generation_max_length: 225
Downloads last month
185
Safetensors
Model size
0.2B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for thunderboltc/gemini_whisper_sanlish

Finetuned
(3746)
this model