Whispered MLA — experiments zoo
Every study checkpoint behind the Whispered MLA release (specialists, mixes, fairness controls, plus_cross variants, matched-budget pairs), so the study can be verified or extended without retraining. Research artifacts — read the warnings.
Load any entry: AutoModelForSpeechSeq2Seq.from_pretrained('burakaydinofficial/whispered-mla-experiments', subfolder='<name>', trust_remote_code=True) (transformers==4.46.x; scheme=none entries load without trust_remote_code).
Warnings: single-language slate/* models DESTROY other languages (that finding is their purpose); *_tiny/_base are 8k-budget concept runs, superseded by the released family; quality claims live in the released models' cards, not here.
| entry | scheme | scope | languages | steps | base |
|---|---|---|---|---|---|
slate_en |
partial_preservation | decoder_self | en | 2000 | whisper-small |
slate_de |
partial_preservation | decoder_self | de | 2000 | whisper-small |
slate_ru |
partial_preservation | decoder_self | ru | 2000 | whisper-small |
slate_ar |
partial_preservation | decoder_self | ar | 2000 | whisper-small |
slate_ka |
partial_preservation | decoder_self | ka | 2000 | whisper-small |
slate_zh-CN |
partial_preservation | decoder_self | zh-CN | 2000 | whisper-small |
fair_en |
none | decoder_self | en | 2000 | whisper-small |
fair_ar |
none | decoder_self | ar | 2000 | whisper-small |
fair_zh-CN |
none | decoder_self | zh-CN | 2000 | whisper-small |
mix_mla |
partial_preservation | decoder_self | en+de+ru+ar+ka+zh-CN | 6000 | whisper-small |
mix_none |
none | decoder_self | en+de+ru+ar+ka+zh-CN | 6000 | whisper-small |
mix11_mla |
partial_preservation | decoder_self | ar+cy+de+en+es+fr+ka+ru+th+tr+zh-CN | 6000 | whisper-small |
mix11_none |
none | decoder_self | ar+cy+de+en+es+fr+ka+ru+th+tr+zh-CN | 6000 | whisper-small |
pxc_en |
partial_preservation | plus_cross | en | 2000 | whisper-small |
pxc_mix11 |
partial_preservation | plus_cross | ar+cy+de+en+es+fr+ka+ru+th+tr+zh-CN | 6000 | whisper-small |
mix11_mla_tiny |
partial_preservation | decoder_self | ar+cy+de+en+es+fr+ka+ru+th+tr+zh-CN | 6000 | whisper-tiny |
mix11_none_tiny |
none | decoder_self | ar+cy+de+en+es+fr+ka+ru+th+tr+zh-CN | 6000 | whisper-tiny |
mix11_mla_base |
partial_preservation | decoder_self | ar+cy+de+en+es+fr+ka+ru+th+tr+zh-CN | 6000 | whisper-base |
mix11_none_base |
none | decoder_self | ar+cy+de+en+es+fr+ka+ru+th+tr+zh-CN | 6000 | whisper-base |
Full per-model provenance: registry.json in this repo.