Wav2Vec2 Base 960h — BonfyreFPQ v12
Native .fpq v12 quantized weights for facebook/wav2vec2-base-960h.
Model Info
| Property | Value |
|---|---|
| Architecture | Speech Recognition (ASR) |
| Parameters | 95M |
| Original Size | 360 MB (safetensors) |
| FPQ v12 Size | 90 MB |
| Compression | 4.0× |
| Avg Cosine | 0.999772 |
| Worst Cosine | 0.998678 |
Format
BonfyreFPQ v12 native format: rANS entropy-coded E8 lattice coordinates with 6-bit packed tiles and FP16 scales.
Decode to safetensors
bonfyre-fpqx decode wav2vec2-base-960h-v12-fpq.fpq output.safetensors
Source
Converted from facebook/wav2vec2-base-960h using BonfyreFPQ v12.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support