Rethinking Classifier-Free Guidance in On-Policy Diffusion Distillation
実装難易度
Easy
推論・学習コスト
High
想定用途
分類
概要
On-policy distillation (OPD) adapts diffusion models by querying a teacher along trajectories generated by the current student, but how it should behave under classifier-free guidance (CFG), a default component of modern diffusion systems, remains poorly understood. Existing OPD methods naturally extend velocity matching to the CFG-composed prediction, directly matching teacher and student guided…
何が新しいか
On-policy distillation (OPD) adapts diffusion models by querying a teacher along trajectories generated by the current student, but how it should behave under classifier-free guidance (CFG), a…
何に使えるか
分類
実装情報
- Hugging Face URL
- あり
実装チェックリスト
実装または配布ページ
OKコードまたはモデル配布ページから検証を始められます。
一次情報リンク
OKHugging Face
検証しやすさ
OK実装またはモデル配布ページから試せる可能性が高いです。
計算資源
要確認学習や高解像度推論ではGPUメモリと実行時間に注意が必要です。
ライセンス
未取得配布元のLICENSE、モデルカード、Paperの利用条件を確認してください。
商用利用
未取得研究利用限定、データセット由来制限、API規約の有無を確認してください。
自社データで試すなら
製造業・材料開発のExcel/CSVデータに落とし込むための最初の手順です。
- 1まず自社データを、入力条件、目的変数、評価したい指標に分けて整理します。
- 2LightGBMやRandom Forestなどのベースラインを先に作り、この手法と比較します。
- 3評価指標はR2/RMSE、AUC、異常検知の再現率、実験回数削減率など、現場の意思決定に近いものを選びます。
- 4SHAPや特徴量重要度で、効いている因子が物理・化学・工程知識と矛盾しないか確認します。
実装難易度
Easy - 実装またはモデル配布ページから試せる可能性が高いです。
必要リソース
- GPU目安: High
- データセット: 論文・リポジトリ側の指定を確認してください。
- 学習要否: 推論だけで試せる可能性があります。
- 学習や高解像度推論ではGPUメモリと実行時間に注意が必要です。
実務で使う場合の注意点
- ライセンスと商用利用条件は、Paper / GitHub / Hugging Face の配布元で確認してください。
- 精度、再現性、計算コストはデータセットや評価条件に依存します。
- 個人情報や機密データを扱う場合は、入力データの保存先と外部API利用条件を確認してください。
関連記事
Shared Semantic Codebook Distillation for Unpaired Cross-Modal Medical Classification
Cross-modal knowledge distillation can transfer diagnostic knowledge from a strong but costly teacher modality
HeteroPROPMT: A Real-time and Privacy-Preserving Heterogeneous Collaborative Perception Framework
Collaborative Perception (CP) improves autonomous systems' awareness of their surroundings by sharing sensor d
Spectral Prior for Reducing Exposure Bias in Diffusion Models
Diffusion models typically suffer from error accumulation during iterative sampling, commonly referred to as e
aarambh-studio — 🦀 Decoder-only LLM built from scratch in pure Rust using Candle — no Python, no PyTorch. Gated DeltaNet + sparse attention, fine-grained MoE, native video/document understanding, long-horizon tool agents, quantization-aware training. Scales: Tiny (25M) to Large (1.3B).
Rust言語でCandleライブラリを利用して、PythonやPyTorchを使用せずにDecoder-only LLMを自作した。