Views
No views yet
openai/whisper-base (unmodified architecture) fine-tuned on the 24 languages of the CC0
Whispered corpus. This is the
matched scientific control for burakaydinofficial/whisper-base-mla-24lang —
trained identically, minus the MHA→MLA conversion — published so the MLA conversion cost is
independently reproducible. No custom code: loads directly in transformers, and — being a plain unmodified Whisper — is convertible for faster-whisper / CTranslate2 / whisper.cpp via their standard converters.1from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor
2model = AutoModelForSpeechSeq2Seq.from_pretrained("burakaydinofficial/whisper-base-24lang") # no trust_remote_code
3processor = AutoProcessor.from_pretrained("burakaydinofficial/whisper-base-24lang")whisper-base-mla-24lang on CommonVoice-17 (scripts/validate.py in the
code repo); the per-language difference is the
conversion cost reported on the MLA card and in docs/results/.Reconstructed numbers. This control's per-language WER/CER in the table are reconstructed as (the MLA model's absolute score − the measured paired conversion cost), not from a separate evaluation of this checkpoint; an independent re-eval may differ by a few tenths. The paired conversion cost is the directly measured quantity.
| Lang | this control |
|---|---|
| en | 20.5 WER |
| de | 29.7 WER |
| es | 18.8 WER |
| fr | 32.9 WER |
| it | 30.5 WER |
| pt | 28.4 WER |
| ru | 28.0 WER |
| nl | 26.9 WER |
| pl | 34.4 WER |
| id | 38.6 WER |
| tr | 38.8 WER |
| hi | 36.3 WER |
| ms | 35.0 WER |
| sv-SE | 43.2 WER |
| th | 23.9 CER |
| zh-CN | 25.2 CER |
| cs | 53.2 WER |
| vi | 43.1 WER |
| fi | 51.0 WER |
| el | 51.6 WER |
| da | 56.0 WER |
| ja | 34.3 CER |
| nn-NO | 67.3 WER |
| ko | 62.2 WER |