visualears-fastconformer-fa-full-ab-nvfp4

NVFP4 (W4A4) post-training quantization of Reza2kn/visualears-fastconformer-fa-full-ab via NVIDIA modelopt.

• Base architecture: EncDecHybridRNNTCTCBPEModel (NeMo)
• Calibration: 32 Persian clips from Reza2kn/persian-asr-eval-v0 (held out from eval).
• Hardware target: NVIDIA Blackwell tensor cores.

Eval — Reza2kn/persian-asr-eval-v0 (FLEURS-fa, 200 clips)

Variant	WER ↓	CER ↓	per-clip latency	peak VRAM
NVFP4 (this repo)	20.33%	7.38%	48 ms	603 MiB

Usage

import nemo.collections.asr as nemo_asr
m = nemo_asr.models.ASRModel.restore_from("visualears-fastconformer-fa-full-ab-NVFP4.nemo").cuda().eval()
transcripts = m.transcribe(["clip.wav"])
print(transcripts[0])

License

Inherits the base model's license.