MotionDetailCache on MAGI-1 (dev4-detail)
基于 FlowCache4MAGI-1-dev3-motion 的扩展分支,在 MotionCache 的 motion 度量 之外,增加 spatial detail 度量,Phase 2 active token 由两者共同决定。
与 dev3 (MotionCache) 的差异
| 维度 | dev3 MotionCache | dev4 MotionDetailCache |
|---|---|---|
| Phase 2 权重 | 仅 motion 权重 W_m | motion W_m + detail W_d 融合 |
| Detail proxy | — | latent 局部 k×k 空间方差 |
| 适用场景 | 运动区域优先计算 | 静止纹理/边缘/语义细节也优先保留 |
Detail 度量
对每个 latent 像素 (t, h, w):
- 通道聚合幅度:
mag = mean(|x|, dim=C)→[N, T, H, W] - 局部空间方差:在
detail_window_size×detail_window_size窗口内计算Var(mag) - Soft-mapping:帧内 min-max 归一化 →
W_d ∈ [detail_alpha, 1]
高方差区域对应边缘、纹理、物体边界等 heterogeneous 邻域。
Motion + Detail 融合
W_combined = combine(W_motion, W_detail)
A[p] += W_combined[p] · Δ_chunk
active[p] = A[p] > τ
weight_combine_mode(默认 max):
| 模式 | 公式 | 特点 |
|---|---|---|
max |
max(W_m, W_d) | 运动或细节任一显著即可加速累积(推荐) |
product |
W_m × W_d | 两者同时高才快速激活,更激进 reuse |
blend |
(1-λ)W_m + λW_d | 线性权衡,λ=detail_lambda |
默认超参
继承 dev3 的 motion 参数,并新增:
| 参数 | 默认值 | 说明 |
|---|---|---|
detail_alpha |
0.5 | detail 权重下限 |
detail_window_size |
3 | 局部方差窗口(奇数) |
detail_lambda |
0.5 | blend 模式下的 detail 权重 |
weight_combine_mode |
max | 融合方式 |
快速运行
cd FlowCache4MAGI-1-dev4-detail
conda activate magi
export CUDA_VISIBLE_DEVICES=0
export MASTER_ADDR=localhost
bash scripts/single_run/motiondetail_t2v.sh
代码结构
inference/pipeline/cache/
├── motioncache.py # dev3 基类 MotionWiseCache
└── motiondetailcache.py # dev4 MotionDetailCache
yaml_config/single_run/motiondetail_config.yaml
scripts/single_run/motiondetail_t2v.sh
其余 sparse forward、KV cache、integrate 逻辑与 dev3 完全一致。