waxal-w2vbert-linsna-seed43 is a Lingala and Shona CTC model, and it anchors six of the seven ensembles that the scored submission selects across; the seventh, p2n_ens_s46anchor, is anchored by the seed-46 arm. It is a fresh fine-tune of facebook/w2v-bert-2.0, a 600M parameter encoder, at seed 43, with its own raw character vocabulary built from unnormalised transcripts, capitals and the punctuation CER scores included.

The word "fresh" is the point. This checkpoint does not continue from waxal-w2vbert-linsnalug-raw; it starts again from the public base with a different seed, so its errors are decorrelated from that entire lineage. Fresh seeds were the most valuable ensemble members measured in this project, worth about +0.0032 each, and were cheaper and more effective than adding new architectures.

Its role as anchor matters more than its solo score. In character-level ROVER the anchor supplies the skeleton, and its text survives every slot unless the members outvote it, so anchor choice moves the result more than any single member does. Using the stronger distilled arm, waxal-w2vbert-linsna-distilled, as the anchor instead scored 0.763419, below the 0.764915 obtained with this checkpoint anchoring and the distilled arm voting as a member.

Training data

source	rows	hours	licence
google/WaxalNLP, Lingala and Shona train and validation	33,002	about 179	CC-BY-4.0
google/WaxalNLP, Lingala and Shona test split	3,615 (1,866 lin, 1,749 sna)	19.2	CC-BY-4.0

30,987 rows after the speaker-disjoint holdout and the validation rows are removed.

Disclosure. The WAXAL Phase 1 test split is used as training data here. Phase 2 permits this and the organisers confirmed it publicly. It is gold reference text released by the organisers, not pseudo-labels, and it is not the Phase 2 test set.

Training configuration

Settings:

setting	value
starting weights	`facebook/w2v-bert-2.0`
seed	43
vocabulary	built fresh from unnormalised transcripts, 97 tokens
epochs	8
learning rate	1.0e-4
warmup ratio	0.1
batch	12 per device, 3 gradient-accumulation steps (effective 36)
precision	bf16, gradient checkpointing on
augmentation	on, applied in the collator

Trained with add_adapter: true and no vocab_dir. Because the vocabulary is rebuilt from these transcripts, this checkpoint's CTC token order differs from the waxal-w2vbert-linsnalug-raw lineage and it is not weight-average compatible with it. It combines by voting, not by averaging. Target normalization is NFC and whitespace collapse only: no lowercasing, no punctuation stripping, because CER is computed on the raw string and WER does not strip punctuation. Hardware: 1x NVIDIA RTX 5090 (32 GB VRAM), 32 CPU cores, 92 GB RAM.

Augmentation is applied on the fly in the data collator, so each epoch sees a fresh draw of the same clip:

MUSAN noise (OpenSLR 17, noise/ subset only, not speech), p = 0.5, SNR 5 to 20 dB
simulated room impulse responses (OpenSLR 28), p = 0.3
Gaussian noise p = 0.1, random gain p = 0.3
pitch and speed perturbation disabled (p = 0.0)

Evaluation

split	error, 0.5 * (WER + CER)
holdout, Lingala and Shona combined	0.2862

The holdout is a speaker-disjoint carve of the WAXAL training data. It is not the competition test set, and it shares recording conditions and language variety with the training corpus, so it flatters any arm tuned on that corpus and gives no independent read on generalisation. Lower is better; the competition score is 1 - 0.5*(WER + CER).

No solo public-leaderboard score was recorded for this checkpoint. Holdout and leaderboard did not always agree in this project, so the holdout figure should be read as a training diagnostic, not as a ranking.

Intended use and limitations

Intended use is automatic speech recognition for Lingala and Shona photo-description speech: spontaneous descriptions of images, on consumer recording equipment, averaging 20.2 seconds per clip. It has not seen Luganda; for Luganda, start from waxal-w2vbert-linsnalug-raw, the only published checkpoint in this family that did. The competition targets Lingala, Shona and Luganda, and the corrected Phase 2 test set was measured to contain only the first two.

This model is one component of the ensembles behind the submission. The 26-member character-level ROVER core scored 0.764915, and the scored final submission, recipe p2n_mbr, selects per clip across seven such ensembles and scored 0.766791 public and 0.772552 private. Solo strength is not the only criterion for membership in those core ensembles. Some members are deliberately weak alone, notably the blank-penalty re-decodes, which over-generate words and lose score in isolation (0.743 against 0.746 for the same checkpoint) while contributing +0.0104 in total, because a character vote can filter a spurious word and can never recover a missing one. The limit of that argument was also measured: genuinely weak checkpoints did not help, and a 21-member vote including them scored 0.764152 against 0.764759 for a 25-member vote of strong sources.

How to load

python

import torch
import soundfile as sf
from transformers import Wav2Vec2BertProcessor, Wav2Vec2BertForCTC

MODEL_ID = "anyantudre/waxal-w2vbert-linsna-seed43"   # or a local checkpoint directory

processor = Wav2Vec2BertProcessor.from_pretrained(MODEL_ID)
model = Wav2Vec2BertForCTC.from_pretrained(MODEL_ID, torch_dtype=torch.float32).eval()

wav, sr = sf.read("clip.wav")   # mono, resample to 16 kHz before this call
inputs = processor(wav, sampling_rate=16000, return_tensors="pt")
with torch.no_grad():
    logits = model(**inputs).logits

print(processor.batch_decode(logits.argmax(-1))[0])

When transcribing a batch rather than a single clip, pass the attention mask through to the model. Clips in a batch have different lengths and the shorter ones are zero-padded; without the mask, self-attention treats that padding as real audio and the encoder output changes for every frame, not just the padded tail. Omitting it altered roughly half of all transcripts in a side-by-side check.