VisualEars FastConformer Persian ASR ONNX W4

Derivative export of Reza2kn/visualears-fastconformer-fa-full-ab.

Artifact

  • Format: ONNX Runtime weight-only 4-bit export
  • Quantization/conversion: ONNX Runtime MatMulNBits, 4-bit asymmetric, block size 32
  • Runtime validation: ONNX Runtime CPU
  • Validation result: 98.61% CTC argmax parity
  • Size: 141 MB, 30.7% of ONNX FP source

Validation

Runtime parity was checked against PyTorch CTC logits on 16 calibration clips padded to 2005 mel frames. The metric is CTC argmax token agreement versus the PyTorch reference logits, not end-to-end WER.

Usage Boundary

These are fixed-frame acoustic CTC-core exports. They take precomputed log-mel features as processed_signal; they are not full raw-audio-to-text pipelines by themselves.

Notes

Best verified ONNX 4-bit export from the fixed CTC core.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Reza2kn/visualears-fastconformer-fa-full-ab-onnx-w4