Whisper Tiny Grpo 0029b82 Cv22 Tiny Sft

---

library_name: transformers license: mit pipeline_tag: automatic-speech-recognition base_model: openai/whisper-tiny datasets:

• cv22_index language:
• en tags:
• whisper
• grpo
• reinforcement-learning
• asr metrics:
• wer
• cer model-index:
• name: whisper-tiny-grpo-0029b82-cv22-tiny-sft results:
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: af split: validation metrics:
    • type: wer value: 0.9111 name: WER (af)
    • type: cer value: 0.3019 name: CER (af)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: am split: validation metrics:
    • type: wer value: 0.8649 name: WER (am)
    • type: cer value: 0.5480 name: CER (am)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ar split: validation metrics:
    • type: wer value: 0.6471 name: WER (ar)
    • type: cer value: 0.2526 name: CER (ar)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: as split: validation metrics:
    • type: wer value: 0.3438 name: WER (as)
    • type: cer value: 0.1929 name: CER (as)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: az split: validation metrics:
    • type: wer value: 0.7442 name: WER (az)
    • type: cer value: 0.2567 name: CER (az)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ba split: validation metrics:
    • type: wer value: 0.8125 name: WER (ba)
    • type: cer value: 0.2775 name: CER (ba)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: be split: validation metrics:
    • type: wer value: 0.5556 name: WER (be)
    • type: cer value: 0.1412 name: CER (be)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: bg split: validation metrics:
    • type: wer value: 0.6000 name: WER (bg)
    • type: cer value: 0.1653 name: CER (bg)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: bn split: validation metrics:
    • type: wer value: 0.3750 name: WER (bn)
    • type: cer value: 0.2233 name: CER (bn)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: br split: validation metrics:
    • type: wer value: 0.7073 name: WER (br)
    • type: cer value: 0.3492 name: CER (br)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ca split: validation metrics:
    • type: wer value: 0.8200 name: WER (ca)
    • type: cer value: 0.3684 name: CER (ca)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: cs split: validation metrics:
    • type: wer value: 0.6250 name: WER (cs)
    • type: cer value: 0.1302 name: CER (cs)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: cy split: validation metrics:
    • type: wer value: 0.4500 name: WER (cy)
    • type: cer value: 0.1422 name: CER (cy)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: da split: validation metrics:
    • type: wer value: 0.3462 name: WER (da)
    • type: cer value: 0.1318 name: CER (da)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: de split: validation metrics:
    • type: wer value: 0.2245 name: WER (de)
    • type: cer value: 0.0888 name: CER (de)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: el split: validation metrics:
    • type: wer value: 0.4091 name: WER (el)
    • type: cer value: 0.1304 name: CER (el)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: en split: validation metrics:
    • type: wer value: 0.5122 name: WER (en)
    • type: cer value: 0.2673 name: CER (en)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: es split: validation metrics:
    • type: wer value: 0.6522 name: WER (es)
    • type: cer value: 0.2416 name: CER (es)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: et split: validation metrics:
    • type: wer value: 0.9630 name: WER (et)
    • type: cer value: 0.2326 name: CER (et)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: eu split: validation metrics:
    • type: wer value: 0.5833 name: WER (eu)
    • type: cer value: 0.1421 name: CER (eu)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: fa split: validation metrics:
    • type: wer value: 0.9000 name: WER (fa)
    • type: cer value: 0.3605 name: CER (fa)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: fi split: validation metrics:
    • type: wer value: 0.5000 name: WER (fi)
    • type: cer value: 0.1276 name: CER (fi)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: fr split: validation metrics:
    • type: wer value: 0.4000 name: WER (fr)
    • type: cer value: 0.1674 name: CER (fr)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: gl split: validation metrics:
    • type: wer value: 0.2812 name: WER (gl)
    • type: cer value: 0.0478 name: CER (gl)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ha split: validation metrics:
    • type: wer value: 0.5161 name: WER (ha)
    • type: cer value: 0.1083 name: CER (ha)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: he split: validation metrics:
    • type: wer value: 1.0000 name: WER (he)
    • type: cer value: 0.6696 name: CER (he)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: hi split: validation metrics:
    • type: wer value: 0.3019 name: WER (hi)
    • type: cer value: 0.1707 name: CER (hi)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ht split: validation metrics:
    • type: wer value: 0.9524 name: WER (ht)
    • type: cer value: 0.4111 name: CER (ht)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: hu split: validation metrics:
    • type: wer value: 1.0556 name: WER (hu)
    • type: cer value: 0.7203 name: CER (hu)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: hy-AM split: validation metrics:
    • type: wer value: 0.8800 name: WER (hy-AM)
    • type: cer value: 0.2772 name: CER (hy-AM)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: id split: validation metrics:
    • type: wer value: 0.7692 name: WER (id)
    • type: cer value: 0.3378 name: CER (id)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: is split: validation metrics:
    • type: wer value: 0.9677 name: WER (is)
    • type: cer value: 0.5430 name: CER (is)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: it split: validation metrics:
    • type: wer value: 0.4211 name: WER (it)
    • type: cer value: 0.1074 name: CER (it)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ja split: validation metrics:
    • type: wer value: 1.0000 name: WER (ja)
    • type: cer value: 0.4375 name: CER (ja)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ka split: validation metrics:
    • type: wer value: 0.7619 name: WER (ka)
    • type: cer value: 0.3833 name: CER (ka)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: kk split: validation metrics:
    • type: wer value: 0.6500 name: WER (kk)
    • type: cer value: 0.2243 name: CER (kk)
  • task: type: automatic-speech-recognition name: Automatic Speech Recognition dataset: type: cv22_index name: cv22_index config: ko split: validation...