Automatic Speech Recognition
Transformers
Safetensors
Sediq
whisper
formosanbank
formosan
endangered-languages
speech
asr
whisper-small
tacl
Instructions to use FormosanBank/formosan-asr-taroko with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use FormosanBank/formosan-asr-taroko with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="FormosanBank/formosan-asr-taroko")# Load model directly from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq processor = AutoProcessor.from_pretrained("FormosanBank/formosan-asr-taroko") model = AutoModelForSpeechSeq2Seq.from_pretrained("FormosanBank/formosan-asr-taroko", device_map="auto") - Notebooks
- Google Colab
- Kaggle
| { | |
| "aggregate": { | |
| "cer_normalized_micro": 0.025800467225505015, | |
| "character_errors": 751, | |
| "corpora": 4, | |
| "reference_characters": 29108, | |
| "reference_words": 6991, | |
| "utterances": 962, | |
| "wer_normalized_micro": 0.060506365326848804, | |
| "word_errors": 423 | |
| }, | |
| "by_corpus": [ | |
| { | |
| "cer_normalized": "0.026968", | |
| "cer_raw": "0.026968", | |
| "char_deletions": "96", | |
| "char_deletions_raw": "96", | |
| "char_insertions": "97", | |
| "char_insertions_raw": "97", | |
| "char_substitutions": "130", | |
| "char_substitutions_raw": "130", | |
| "eval_corpus": "ILRDF", | |
| "eval_language": "taroko", | |
| "eval_utterances": "474", | |
| "reference_characters": "11977", | |
| "reference_characters_raw": "11977", | |
| "reference_words": "2881", | |
| "reference_words_raw": "2881", | |
| "run_id": "whisper_small_pan_formosan_to_language_train-taroko_eval-taroko-all-corpora_dataset_v1_split_v2_formosan_safe_v1_seed13", | |
| "wer_normalized": "0.083652", | |
| "wer_raw": "0.083652", | |
| "word_deletions": "12", | |
| "word_deletions_raw": "12", | |
| "word_insertions": "17", | |
| "word_insertions_raw": "17", | |
| "word_substitutions": "212", | |
| "word_substitutions_raw": "212" | |
| }, | |
| { | |
| "cer_normalized": "0.046866", | |
| "cer_raw": "0.046866", | |
| "char_deletions": "38", | |
| "char_deletions_raw": "38", | |
| "char_insertions": "26", | |
| "char_insertions_raw": "26", | |
| "char_substitutions": "22", | |
| "char_substitutions_raw": "22", | |
| "eval_corpus": "ePark1", | |
| "eval_language": "taroko", | |
| "eval_utterances": "125", | |
| "reference_characters": "1835", | |
| "reference_characters_raw": "1835", | |
| "reference_words": "420", | |
| "reference_words_raw": "420", | |
| "run_id": "whisper_small_pan_formosan_to_language_train-taroko_eval-taroko-all-corpora_dataset_v1_split_v2_formosan_safe_v1_seed13", | |
| "wer_normalized": "0.090476", | |
| "wer_raw": "0.090476", | |
| "word_deletions": "4", | |
| "word_deletions_raw": "4", | |
| "word_insertions": "1", | |
| "word_insertions_raw": "1", | |
| "word_substitutions": "33", | |
| "word_substitutions_raw": "33" | |
| }, | |
| { | |
| "cer_normalized": "0.073411", | |
| "cer_raw": "0.073411", | |
| "char_deletions": "2", | |
| "char_deletions_raw": "2", | |
| "char_insertions": "93", | |
| "char_insertions_raw": "93", | |
| "char_substitutions": "69", | |
| "char_substitutions_raw": "69", | |
| "eval_corpus": "ePark2", | |
| "eval_language": "taroko", | |
| "eval_utterances": "58", | |
| "reference_characters": "2234", | |
| "reference_characters_raw": "2234", | |
| "reference_words": "531", | |
| "reference_words_raw": "531", | |
| "run_id": "whisper_small_pan_formosan_to_language_train-taroko_eval-taroko-all-corpora_dataset_v1_split_v2_formosan_safe_v1_seed13", | |
| "wer_normalized": "0.043315", | |
| "wer_raw": "0.043315", | |
| "word_deletions": "12", | |
| "word_deletions_raw": "12", | |
| "word_insertions": "2", | |
| "word_insertions_raw": "2", | |
| "word_substitutions": "9", | |
| "word_substitutions_raw": "9" | |
| }, | |
| { | |
| "cer_normalized": "0.013627", | |
| "cer_raw": "0.013627", | |
| "char_deletions": "62", | |
| "char_deletions_raw": "62", | |
| "char_insertions": "39", | |
| "char_insertions_raw": "39", | |
| "char_substitutions": "77", | |
| "char_substitutions_raw": "77", | |
| "eval_corpus": "ePark3", | |
| "eval_language": "taroko", | |
| "eval_utterances": "305", | |
| "reference_characters": "13062", | |
| "reference_characters_raw": "13062", | |
| "reference_words": "3159", | |
| "reference_words_raw": "3159", | |
| "run_id": "whisper_small_pan_formosan_to_language_train-taroko_eval-taroko-all-corpora_dataset_v1_split_v2_formosan_safe_v1_seed13", | |
| "wer_normalized": "0.038303", | |
| "wer_raw": "0.038303", | |
| "word_deletions": "9", | |
| "word_deletions_raw": "9", | |
| "word_insertions": "9", | |
| "word_insertions_raw": "9", | |
| "word_substitutions": "103", | |
| "word_substitutions_raw": "103" | |
| } | |
| ], | |
| "condition": "pan_formosan_to_language", | |
| "dataset_version": "dataset_v1", | |
| "iso_639_3": "trv", | |
| "language": "Taroko", | |
| "language_id": "taroko", | |
| "model_family": "whisper_small", | |
| "normalization_version": "formosan_safe_v1", | |
| "readiness": { | |
| "actual_train_rows": 7650, | |
| "best_selection_cer": 0.026846856509768723, | |
| "best_selection_wer": 0.061902180140190874, | |
| "best_step": 1722, | |
| "condition": "pan_formosan_to_language", | |
| "dominant_hypothesis_rate": 0.005197505197505198, | |
| "estimated_effective_epochs": 3.0002614379084966, | |
| "failures": 0, | |
| "generated_at": "2026-06-25T21:34:57+00:00", | |
| "intended_train_rows": 7650, | |
| "manifest_dir": "/projects/prudlab/tacl_formosan_asr/experiments/manifests/gate5_whisper_final_v1_with_pan_checkpoint/whisper_small_pan_formosan_to_language_train-taroko_eval-taroko-all-corpora_dataset_v1_split_v2_formosan_safe_v1_seed13", | |
| "nonempty_hypothesis_rate": 1.0, | |
| "prediction_rows": 962, | |
| "result_dir": "/scratch/scheppat/projects/tacl_formosan_asr/runs_whisper_final_v1/whisper_small_pan_formosan_to_language_train-taroko_eval-taroko-all-corpora_dataset_v1_split_v2_formosan_safe_v1_seed13", | |
| "result_label": "final", | |
| "run_id": "whisper_small_pan_formosan_to_language_train-taroko_eval-taroko-all-corpora_dataset_v1_split_v2_formosan_safe_v1_seed13", | |
| "status": "pass", | |
| "steps": 2869, | |
| "test_rows": 962, | |
| "train_batch_size": 8, | |
| "warnings": 0 | |
| }, | |
| "seed": 13, | |
| "split_version": "split_v2" | |
| } | |