Downloads · 30 days
8
13% of all-time downloads
affectscore-ace-step-r32-20260629-no-style
HiiragiLee/affectscore-ace-step-r32-20260629-no-style
affectscore-ace-step-r32-20260629-no-style is a machine learning model from HiiragiLee. 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 peft. The card lists the license as mit.
Ablation checkpoint for the AffectScore rank sweep. Identical to the full r=32 model except Layer 2 style-text descriptor is removed from the conditioning prompt. The Layer 1 V-A mood-word prompt is retained. CLAP aux…
All-time downloads
61
Public
Repo size
31.5 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors31.5 MB · 100%
At a glance
- Library
- peft
- License
- mit
- Access
- Public
- Created
- Jun 30, 2026
- Updated
- Aug 7, 2026
- SHA
- c84757a8
Base models
Model card
From the Hugging Face model README
AffectScore -- LoRA r=32, no-style (ablation)
Ablation checkpoint for the AffectScore rank sweep. Identical to the full r=32 model except Layer 2 style-text descriptor is removed from the conditioning prompt. The Layer 1 V-A mood-word prompt is retained. CLAP auxiliary monitoring is additionally disabled (redundant since L_CLAP carries zero gradient to LoRA weights regardless of λ, but disabled for consistency).
This checkpoint isolates the contribution of the Layer 2 style-text descriptor. It is not intended for production use.
Code and Colab notebooks: github.com/LeeTgk/affectscore (link active after publication)
Training
| Parameter | Value |
|---|---|
| Base model | ACE-Step/ACE-Step-v1-3.5B |
| LoRA rank | 32 |
| LoRA targets | cross_attn.q_proj, cross_attn.v_proj |
| Layer 1 (V-A mood words) | enabled |
| Layer 2 (style-text descriptor) | disabled |
| λ_CLAP | 0 (monitoring disabled) |
| Training clips | 1,571 (CC0/CC-BY, CLAP-quality filtered) |
| Epochs | 50 |
| Learning rate | 1e-4 |
| Batch size | 8 |
| Hardware | NVIDIA A100 (Google Colab) |
Training dataset: https://zenodo.org/records/21830658
Evaluation
All metrics computed on 200 held-out clips at 40 diffusion steps.
Audio quality (EVAL-02)
| FAD-MERT ↓ | CLAP-score | KLD ↓ | PCE ↑ |
|---|---|---|---|
| 32.57 | n/a | 1.12 | 3.56 |
CLAP-score is undefined for this variant (no style-text reference string).
Training loss: L_diff = 1.1713 at epoch 50 (Δ = +0.0047 above full r=32).
Ablation variants
See affectscore-ace-step-r32-20260629 for the full variant table and evaluation discussion.
Citation
Citation will be added once the paper is published. If you use this model before then, please link to this repository.
- Library
- peft
- License
- mit
- Created
- Jun 30, 2026
- Updated
- Aug 7, 2026