Downloads · 30 days
0
ffasr-bd6-rank32-adapter-weights
kienpt0901/ffasr-bd6-rank32-adapter-weights
ffasr-bd6-rank32-adapter-weights is a machine learning model from kienpt0901. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as cc-by-4.0.
LoRA adapter for efficient-speech/lite-whisper-large-v3-turbo-acc (revision ef2c0dd768cc9832a8a5a3397ab7218c838fea66), trained for FFASR robustness.
Access
Public
Updated Jul 18, 2026
Repo size
13.8 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors13.8 MB · 100%
At a glance
- License
- cc-by-4.0
- Access
- Public
- Created
- Jul 18, 2026
- Updated
- Jul 18, 2026
- SHA
- a3b3fb7a
Base models
Model card
From the Hugging Face model README
FFASR robustness LoRA adapter (Pilot BD6, rank-32)
LoRA adapter for efficient-speech/lite-whisper-large-v3-turbo-acc
(revision ef2c0dd768cc9832a8a5a3397ab7218c838fea66), trained for FFASR robustness.
- LoRA rank 32, alpha 64, dropout 0 at inference
- Targets:
model.encoder.layers.*.self_attn.{q,v,out}_projandmodel.decoder.layers.*.encoder_attn.{q,v,out}_proj - Processor:
openai/whisper-large-v3 - Trained on 100h LibriSpeech train-clean-100 (251 speakers) with synthetic reverb/noise/spectral augmentation; clean-anchor CE objective. Checkpoint u110.
Weights-only repo for anonymous download by the FFASR evaluation worker.
- License
- cc-by-4.0
- Created
- Jul 18, 2026
- Updated
- Jul 18, 2026