Fine-tuned checkpoint of OmniLingual ASR CTC 300M_v2 on a mixture of FLEURS Chichewa/Nyanja (2,653 examples, ~10.87 hours) and the Chichewa Trigrams speech-text parallel corpus (25,991 examples, ~10.87 hours — 100% of FLEURS train hours). Training used the fairseq2 wav2vec2 ASR recipe for 5,000 steps with the encoder frozen for the first 1,000 steps. Tokenizer: omniASR_tokenizer_written_v2.

Results

Dataset	WER (%)	CER (%)
FLEURS Chichewa test	35.27	7.66
Zambezi Voice Nyanja test	60.79	13.29

This model outperforms the zero-shot omniASR CTC 3B baseline (35.80% WER on FLEURS, 62.10% on Zambezi Voice) despite having 10× fewer parameters. Compared to fine-tuning on FLEURS alone (36.53% FLEURS / 61.93% Zambezi), the trigram augmentation reduces WER by 1.26 pp on FLEURS and 1.14 pp on Zambezi.

Training details

Base model: omniASR_CTC_300M_v2
Tokenizer: omniASR_tokenizer_written_v2
Training data: FLEURS Chichewa train (2,653 examples, 10.87 h) + Chichewa Trigrams (25,991 examples, 10.87 h)
Steps: 5,000 (best checkpoint at step 5,000 — WER still decreasing at end of training)
Dev WER at best checkpoint: 36.92% (FLEURS dev, 305 examples)
Learning rate: 1e-5
Encoder frozen: first 1,000 steps
Gradient accumulation: 4 batches
Mixed precision: bfloat16
Framework: fairseq2

Checkpoint format

The checkpoint is stored in fairseq2's sharded format. checkpoint/model/pp_00/tp_00/sdp_00.pt contains the full model state (training used a single GPU, so there is only one shard).

To load the model, use the OmniLingual ASR inference pipeline with the omniASR_CTC_300M_v2 model card and point the checkpoint path at sdp_00.pt.

ai4good-labyrinth/omniASR-CTC-300M-v2-chichewa-mixed-1p0

omniASR CTC 300M_v2 — fine-tuned on Chichewa/Nyanja (FLEURS + trigram 1:1)

Results

Training details

Checkpoint format

ai4good-labyrinth/
omniASR-CTC-300M-v2-chichewa-mixed-1p0