Downloads · 30 days
0
deepvoice-pjs-v3.1-ad-lora
ssafy19-dream/deepvoice-pjs-v3.1-ad-lora
deepvoice-pjs-v3.1-ad-lora is a machine learning model from ssafy19-dream. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as other.
SSAFY / deepvoicedetecting v3.1 AntiDeepfake LoRA adapter.
Access
Public
Updated Sep 23, 2026
Repo size
3 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt3 MB · 100%
At a glance
- Library
- transformers
- License
- other
- Access
- Public
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026
- SHA
- 8ddef0b3
Base models
Model card
From the Hugging Face model README
deepvoice-pjs-v3.1-ad-lora
SSAFY / deepvoice_detecting v3.1 AntiDeepfake LoRA adapter.
Base
nii-yamagishilab/xls-r-1b-anti-deepfake- Adapter only (
ad_lora.pt). Does not include full XLS-R-1B weights.
Training (summary)
- Target: encoder last 4 layers +
proj_fcLoRA (rank 8, alpha 16) - Aug (v2.5-style): phone 0.12, codec 0.08, band 0, gain/noise 0.1, overlays
- Objective: fake-priority, real_hit floor ~0.90
- Valid (internal): fake_hit ≈ 0.92, real_hit ≈ 0.91 after epoch 3
Usage
Place as model/ad_lora.pt in the contest repo. Inference loads it automatically:
python script.py --ad-lora model/ad_lora.pt --vf-mode antideepfake
# or ensemble with DF-Arena:
python script.py --ad-lora model/ad_lora.pt --vf-mode ensemble
Files
ad_lora.pt— LoRA A/B tensors + meta (adapter: antideepfake_xlsr_lora)
- Library
- transformers
- License
- other
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026