ai4good-labyrinth/fleurs-only-whisper-tiny-no-language

FLEURS-Only Whisper Small No-Language Checkpoint

Summary

This repository contains a Whisper checkpoint for Chichewa/Nyanja automatic speech recognition, fine-tuned from openai/whisper-small.

• Experiment type: fleurs-only
• Base model: openai/whisper-small
• Training condition: no_language
• Release artifact: full fine-tuned checkpoint selected from the best training checkpoint

Intended use

This checkpoint is intended for research and evaluation on Chichewa/Nyanja ASR. It is not a production-ready speech system and should be validated carefully before downstream use.

How to use

This repository contains a full fine-tuned checkpoint. It can be loaded directly with Transformers.

from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor

model_id = "ai4good-labyrinth/fleurs-only-whisper-tiny-no-language"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id)

Training data

• Training source: TBD
• Evaluation source during training: TBD
• Train examples before duration filtering: TBD
• Train examples after duration filtering: TBD
• Dev examples before duration filtering: TBD
• Dev examples after duration filtering: TBD
• Duration filter used during training: min_duration_seconds=TBD, max_duration_seconds=TBD

Training procedure

• Fine-tuning script: experiments/whisper_finetune/finetune_whisper.py
• Base model: openai/whisper-small
• Task: transcribe
• Language hint during training/evaluation: none, corresponding to --language auto in standalone evaluation
• Mixed precision: TBD
• Gradient checkpointing: TBD
• Selected checkpoint step: 1500
• Selected checkpoint epoch: 4.52

Training-time dev selection

The best checkpoint was selected using trainer-side dev evaluation on the duration-filtered FLEURS dev split.

• Dev WER: 0.7683
• Dev CER: 0.2782
• Dev loss: 1.0449

These values come from the training pipeline and may differ slightly from standalone post-hoc evaluation because the decoding path is not perfectly identical.

Evaluation protocol

Standalone evaluation is recommended for the final release. Filtered and unfiltered results should be reported separately.

• Filtered evaluation: min_duration_seconds=0, max_duration_seconds=30
• Unfiltered evaluation: no duration constraint
• Decoding task: transcribe
• Language hint: auto

Evaluation summary

Dataset	Split	Setting	Num examples	WER	CER	Notes
FLEURS	dev	filtered	305	0.8900	0.3060	Filtered to 30 seconds
FLEURS	dev	unfiltered	311	0.8789	0.3023	No duration filter
FLEURS	test	filtered	745	0.8421	0.3248	Filtered to 30 seconds
FLEURS	test	unfiltered	761	0.8343	0.3198	No duration filter
Zambezi	dev	filtered	613	0.9802	0.3922	Filtered to 30 seconds
Zambezi	dev	unfiltered	622	0.9752	0.3882	No duration filter
Zambezi	test	filtered	427	0.8825	0.2860	Filtered to 30 seconds
Zambezi	test	unfiltered	428	0.8792	0.2842	No duration filter

Files in this repository

• Model weights and config: repository root
• Processor/tokenizer files: repository root
• Evaluation JSON files: eval/...

Known limitations

• Whisper does not provide an official Nyanja/Chichewa language token.
• Users must also comply with the upstream dataset licenses and any upstream model license obligations.
• Standalone evaluation and trainer-side evaluation can differ slightly even on the same split and duration filter.
• Cross-dataset results should be interpreted carefully because transcription conventions may differ across corpora.

Citation

If you use this checkpoint, please cite:

• the Whisper paper
• the FLEURS dataset
• this repository

@misc{fleurs_only_whisper_tiny_no_language_2026,
  title        = {FLEURS-Only Whisper Small No-Language Checkpoint},
  author       = {AI4Good Labyrinth Team},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/ai4good-labyrinth/fleurs-only-whisper-tiny-no-language}},
  note         = {Whisper fine-tuning for Chichewa/Nyanja ASR}
}