AIT-86M-GGUF
Quantized GGUF releases for augmem/AIT-86M.
Historical Local Gate Baseline
The earlier card revision dropped the historical local gate summary that was previously kept in the TE-86M release directory. That baseline is restored here for continuity.
Attached JSON:
teacher_dual_mn20whisper_exact_gate_baseline_20260424T155324Z.json
Seeded split-excluded baseline at 1280d:
| Slice | Metric |
|---|---|
| Speech holdout A->T R@1 | 0.5652 |
| Speech holdout T->A R@1 | 0.5992 |
| Speech holdout avg R@1 | 0.5822 |
| WavCaps FSD A->T R@1 | 0.1078 |
| WavCaps FSD T->A R@1 | 0.1030 |
| WavCaps FSD avg R@1 | 0.1054 |
| SALT A->I R@1 | 0.1692 |
| SALT I->A R@1 | 0.1261 |
These figures are reference metrics for the base checkpoint family. The GGUF files here are quantized distribution artifacts and are not separately rebenchmarked in this repo.
Files
| File | Quantization |
|---|---|
AIT-86M-q8_0.gguf |
Q8_0 |
AIT-86M-q5_0.gguf |
Q5_0 |
teacher_dual_mn20whisper_exact_gate_baseline_20260424T155324Z.json |
Restored canonical local gate baseline summary |
These are compact quantized exports for edge deployment and custom runtimes.
They are distribution artifacts for the AIT-86M model family, not a separate model line. Downstream runtime validation remains application-specific.
Full Model
Use the full safetensors checkpoint for training, conversion, or maximum compatibility:
License
Apache 2.0
- Downloads last month
- 185
5-bit
8-bit
Model tree for augmem/AIT-86M-GGUF
Base model
augmem/AIT-86M