- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper investigates how to accelerate inference in multi-stage 3D generative models by optimizing the entire generation pipeline. It introduces two components:
-
PaCO. It jointly searches for compute allocations across stages and sampling schedules within each stage, seeking Pareto-optimal trade-offs between generation quality and computational cost.
-
PSWA. It organizes irregular sparse attention windows into reusable packed attention schedules and caches window boundaries, reducing execution overhead while preserving the original attention semantics.
Critical strengths
-
the studied topic is very important.
-
3D generation is not easy. This paper presents a substantial amount of work.
-
Combining 3D generation and MOO is new(at least for me).
Critical weaknesses
-
The writing styles look like CVPR/ICCV, in ML papers, placing a figure before abstract is not very ususal.
-
Too many bullets, such as make some concluding text bold and put them in front of a para. You'd better to use more words to analyze or tell your methods clearly, not to conclude them too much.
-
line 284, it is unusual to directly put specific numbers into the main method.
-
too many important things are referred to app. This makes the writing feels somewhat disjointed and lacks a smooth flow.
-
pareto compute optimizer is lack of theoretical analysis.
-
The relationship between PACO and PSWA is not very strong. Those two algs. seem pretty independent.
Questions for authors
-
for the lex order, if two var.s are very close, they are deemed as approximately equal, will lex order in your work figure this issue out?
-
consider using some packages in a multiobjective software called libmoon and conduct some baseline experiments.
-
line 317, NSGA2 style search is too old. Can you provide a more advanced MOEA based method?
-
table 1, why only numbers 3, 6, 12 are chosen, whose number is a bit small. How is its general ability.
-
what is the main reason for the efficiency improvement? By algorithm or by implement? is efficiency a major contribution of this work?
-
Is it possible to learn the entire PF/PS, called pareto set/front learning?
-
how is a scaling law against number (m) objectives? is it O(m)?
-
How much more time is needed to find N PO solutions instead of just one? Does it take O(N) times as long?