2.2 KiB
2.2 KiB
Diffusion Acceleration Features
Use with the exact diffusion submodule plus platform, cache, quantization, or benchmarking contracts selected by the live path.
| Feature family | Designs |
|---|---|
| Parallelism | CFG, Expert, HSDP, Pipeline, Sequence, Tensor, VAE patch |
| Attention | Skip-Softmax |
| Quantization | Quantization |
| Cache | Cache-DiT, TeaCache |
| Batching | Diffusion Continuous Batching |
| Offload | Distributed Layerwise Offload |
Feature checks
- Verify feature selection, supported models/platforms/topologies, defaults, dependencies, and every touched feature combination.
- Preserve feature-off and single-rank correctness; reject unsupported combinations before collectives, loading, or execution begin.
- Trace shapes, shards, ranks, collectives, precision, cache validity, residency, and per-request state through the selected runtime hooks.
- Compare output quality or accuracy against the disabled/reference path before accepting latency or memory improvements.
- Require exact commands, topology, environment, warmup, repetitions, and measurements for performance claims; update feature compatibility docs.