mbfa-vietnamese (p4mbfa · vietnamese)

A context-free CNN phoneme encoder and forced aligner (p4mbfa, the successor of CUPE / p3cupe), for the vietnamese language group of standard_g2p. Each 5 ms frame is classified from at most 120 ms of audio (38.9 ms receptive field), so the model cannot learn any language's phonotactics. Phone boundaries come from a segmental Viterbi decoder over the known phone sequence, refined to sub-frame precision at the crossing of neighbouring posteriors.

head	labels	size
ph	local tokens of vietnamese (incl. <blank> SIL noise <unk>)	45
phg	gold phoneme groups, shared by every language group	15
tone	tone values, shared by every language group; trained on this group's tone layer	22

Languages (the FLEURS languages it was trained on): vi. standard_g2p maps every member of the group onto the same tokens, so other members may align too, untested.

Training

Data: FLEURS vietnamese, all 8.7 h, training noise_level 0.02.
Checkpoint: experiment mv01a, epoch 6, selected on FLEURS val_loss (best of 30 epochs; final epoch 3.602, tone CE also worse).
Trunk and phg_head from latin ma02a (final epoch), ph_head and tone_head trained fresh.

metric	value
val_loss	3.3472
val_frame_acc	0.2893
val_frame_acc_groups	0.4653
val_peak_ce_tone	0.879

val_* are on held-out FLEURS clips but are scored against the model's own alignments (there are no boundary labels outside English), so they measure self-consistency, not boundary accuracy.

Labels are standard_g2p dictionary pronunciations (gold inventory 9438371ed6dd), not phonetic transcriptions of what was said.

Usage

The code is in https://github.com/tabahi/bfa_models (p4mbfa/); clone it and run from its root.

from p4mbfa.inference import MbfaAligner

aligner = MbfaAligner.from_pretrained("Tabahi/mbfa-vietnamese")
wav = aligner.load_audio("clip.wav")                     # mono, 16000 Hz
phones = ["SIL", "h", "ɛ", "l", "o", "SIL"]   # this group's tokens (config.json labels.tokens)
for seg in aligner.align(wav, phones):
    print(seg["token"], seg["start_ms"], seg["end_ms"])

aligner.encode(wav) returns the per-frame log posteriors of all three heads. Text -> tokens is standard_g2p's job (goldG2P.phonemize_sentence then lang_group_inventory.to_local); config.json lists the token strings. from_pretrained downloads only config.json and model.safetensors.

Fine-tuning

ckpt/vietnamese_fleurs9h_mv01a_e6_val_loss=3.347.ckpt is the training checkpoint (PyTorch Lightning, weights only, a pickle: load it only if you trust this repo). Download it, then point a p4mbfa yaml at it to continue training, or to train a new language group on its trunk and shared phg head:

hf download Tabahi/mbfa-vietnamese "ckpt/vietnamese_fleurs9h_mv01a_e6_val_loss=3.347.ckpt" --local-dir tmp/hf/mbfa-vietnamese

ckpt_path: "tmp/hf/mbfa-vietnamese/ckpt/vietnamese_fleurs9h_mv01a_e6_val_loss=3.347.ckpt"
reset_fine_heads: true      # new lang_group: rebuild ph_head / tone_head; false = same group

License: AGPL-3.0.