Release best Gemini-tuned Whisper Base and Small with code, normalization and evaluation
cd9b2d8 verified | { | |
| "valid_clips": 65163, | |
| "split_clips": { | |
| "train": 58380, | |
| "test": 3544, | |
| "validation": 3239 | |
| }, | |
| "split_hours": { | |
| "train": 174.0395300111143, | |
| "test": 10.389036091666668, | |
| "validation": 9.740547116666654 | |
| }, | |
| "valid_total_hours": 194.1691132194476, | |
| "language_strings": { | |
| "Russian": 104, | |
| "English": 32223, | |
| "German": 22228, | |
| "Spanish": 2849, | |
| "Hindi": 83, | |
| "German/English": 185, | |
| "French": 2901, | |
| "unclear": 2744, | |
| "Vietnamese": 65, | |
| "Pashto": 1, | |
| "Japanese": 302, | |
| "English / Mandarin Chinese": 1, | |
| "Portuguese": 65, | |
| "Arabic": 92, | |
| "Mandarin Chinese/English": 3, | |
| "German/French/English": 1, | |
| "Malayalam": 24, | |
| "Hungarian": 14, | |
| "Tagalog": 19, | |
| "Swahili": 11, | |
| "English/French": 7, | |
| "en": 99, | |
| "Burmese": 8, | |
| "English/German": 45, | |
| "Romanian": 31, | |
| "Ukrainian": 25, | |
| "Urdu": 44, | |
| "French/German/English": 1, | |
| "Indonesian": 30, | |
| "Bengali": 21, | |
| "English / German": 12, | |
| "Sanskrit": 8, | |
| "German/Mandarin Chinese": 6, | |
| "none": 7, | |
| "Arabic/Thai": 1, | |
| "French/English": 78, | |
| "Gujarati": 8, | |
| "Kinyarwanda": 3, | |
| "Tagalog/English": 2, | |
| "English / Vietnamese": 1, | |
| "Greek": 10, | |
| "Korean": 65, | |
| "Sinhala": 9, | |
| "Nepali": 11, | |
| "English / Tagalog": 1, | |
| "Igbo": 1, | |
| "Haitian Creole": 8, | |
| "Khmer": 24, | |
| "English / Russian": 4, | |
| "Polish/English": 1, | |
| "English/Spanish": 41, | |
| "Hebrew": 14, | |
| "Scots": 7, | |
| "Latvian": 1, | |
| "Russian/English": 13, | |
| "German/Mandarin": 3, | |
| "Swedish": 19, | |
| "Tamil": 29, | |
| "Icelandic": 11, | |
| "Persian": 10, | |
| "Armenian": 4, | |
| "Thai": 38, | |
| "Polish": 25, | |
| "Latin": 3, | |
| "Mandarin Chinese": 68, | |
| "Kannada": 9, | |
| "English/Spanish/Japanese": 2, | |
| "Turkish": 13, | |
| "French/Arabic": 1, | |
| "Dutch": 14, | |
| "Uzbek": 2, | |
| "Luxembourgish": 3, | |
| "Korean/English": 1, | |
| "Cantonese": 10, | |
| "Mandarin": 3, | |
| "Indonesian/English": 1, | |
| "English / Spanish": 18, | |
| "Telugu": 34, | |
| "English/Sanskrit": 2, | |
| "zxx": 11, | |
| "Javanese": 10, | |
| "Danish": 2, | |
| "Macedonian": 1, | |
| "Italian": 33, | |
| "Mandarin Chinese / Dutch": 1, | |
| "German/Japanese": 5, | |
| "Kurdish": 2, | |
| "Italian/English": 3, | |
| "Hawaiian/English": 1, | |
| "English / Malayalam": 1, | |
| "Spanish/English": 14, | |
| "Punjabi": 15, | |
| "Norwegian": 3, | |
| "Mandarin/German": 1, | |
| "Catalan": 1, | |
| "Japanese/English": 6, | |
| "Yiddish": 1, | |
| "English / Japanese": 2, | |
| "Mandarin/English": 1, | |
| "Finnish": 4, | |
| "Azerbaijani": 1, | |
| "Czech": 7, | |
| "German / English": 5, | |
| "Arabic/English": 4, | |
| "Assamese": 3, | |
| "Jamaican Patois": 4, | |
| "Mandarin Chinese/German": 4, | |
| "German/Turkish": 3, | |
| "Slovak": 2, | |
| "Luganda": 2, | |
| "Bulgarian": 6, | |
| "English/Japanese": 2, | |
| "Urdu/English": 1, | |
| "Twi": 5, | |
| "German/Russian": 1, | |
| "Spanish / English": 7, | |
| "Korean/Mongolian": 1, | |
| "French / English": 2, | |
| "Swiss German": 7, | |
| "English/Turkish": 1, | |
| "Marathi": 6, | |
| "English / Italian": 2, | |
| "English/Hindi": 1, | |
| "English/Italian": 5, | |
| "German/Swedish": 1, | |
| "Hindi/English": 3, | |
| "German/Polish": 1, | |
| "Urdu/Arabic": 1, | |
| "English / Korean": 2, | |
| "English / French": 8, | |
| "French/German": 1, | |
| "Vietnamese/English": 2, | |
| "Somali": 4, | |
| "Malay": 1, | |
| "Georgian": 2, | |
| "Arabic / English": 1, | |
| "Mandarin / German": 1, | |
| "Japanese/German": 2, | |
| "English/Pali": 1, | |
| "English/Russian": 2, | |
| "English/Mandarin": 2, | |
| "Serbian": 1, | |
| "Korean / English": 1, | |
| "German/Persian": 1, | |
| "Bhojpuri": 1, | |
| "Portuguese / English": 1, | |
| "Fula / Arabic": 1, | |
| "Sanskrit/Tamil": 1, | |
| "English / Arabic": 2, | |
| "Hindi / English": 1, | |
| "English / Hindi": 1, | |
| "Croatian": 1, | |
| "Sundanese": 1, | |
| "Ukrainian/Russian": 1, | |
| "Moroccan Arabic": 1, | |
| "English / Urdu": 1, | |
| "German/Hungarian": 1, | |
| "Māori/English": 1, | |
| "German/unclear": 1, | |
| "Albanian": 1, | |
| "English / [unclear]": 1, | |
| "Klingon": 1, | |
| "Arabic/Urdu": 1, | |
| "French/Portuguese/English": 1, | |
| "Arabic/Malayalam": 1, | |
| "Scots/Scottish English": 1, | |
| "Isan / Lao": 1, | |
| "Lao": 1, | |
| "German/Arabic": 1, | |
| "Turkish/English": 1, | |
| "Low German": 1, | |
| "Russian/Polish": 1, | |
| "Modern Greek": 1, | |
| "Kazakh": 1, | |
| "multilingual": 1, | |
| "English / Nigerian Pidgin": 1, | |
| "Taiwanese Hokkien": 1, | |
| "Saraiki": 1, | |
| "German/Italian": 1, | |
| "Hokkien": 1, | |
| "English/Portuguese": 1, | |
| "English / Bengali": 1, | |
| "German/Finnish": 1 | |
| }, | |
| "source_membership_counts_nonexclusive": { | |
| "S1_100h": 42701, | |
| "poc": 7579, | |
| "talent": 10917, | |
| "balanced_40x3k": 3979 | |
| }, | |
| "prepared_targets_sha256": "4ebac2622f35d6ce91f7956c0f86c16276d6f45a19e1d1ff19f6eaaace6dfe44", | |
| "deduplication": "Exact selected compressed audio SHA256 across all selected sources/tasks", | |
| "split": "90/5/5 deterministic group hash; explicit speaker or synthetic family before exact audio SHA", | |
| "excluded_clips": 1036, | |
| "captions_used_as_model_inputs": false | |
| } | |