Isotr0py 0c82bb1315 Gate persistent DiT staging on decode-graph models ...

ModelLevelOffloadBackend passed persistent_dit_staging=(device is CUDA) for
every model, so Qwen-Image 2.0/FLUX/WAN-style sequential offload kept DiT
device storage alive across swaps instead of freeing VRAM. Only retain
staging when a DiT actually captures weight pointers (enable_cuda_graph_decode);
other models keep plain move semantics. Add tests pinning both behaviors.

2026-09-22 08:58:32 +00:00

..

conftest.py

[diffusion][feature] Add component-selective offload policies (#5929)

2026-09-06 05:40:09 +08:00

helpers.py

[diffusion][feature] Add component-selective offload policies (#5929)

2026-09-06 05:40:09 +08:00

test_backend_plan_contract.py

[Refactor][Diffusion] Complete generic backend plan consumption (#7313)

2026-09-15 08:40:04 +08:00

test_cuda_host_registration.py

[Diffusion] Register HWR mmap for direct DLO H2D (#6591)

2026-08-25 23:30:11 +08:00

test_diffusion_cpu_offload.py

[Bugfix][Test] Assert stable-audio CPU offload savings on max_memory_allocated (#6826)

2026-09-19 10:04:58 -07:00

test_diffusion_layerwise_offload.py

[CI/Build] Share nightly CUDA YAML between H100/L4 and B200 (#7028)

2026-09-14 14:27:10 +08:00

test_distributed_layerwise_backend.py

[Refactor][Diffusion] Resolve distributed layerwise topology in the plan (#7326)

2026-09-19 19:05:18 +08:00

test_layerwise_backend.py

[Refactor][Diffusion] Complete generic backend plan consumption (#7313)

2026-09-15 08:40:04 +08:00

test_module_collector.py

Extend Cosmos3 CPU offloading to component and layerwise paths (#4695)

2026-07-13 09:48:53 +00:00

test_module_residency.py

[Model] Optimize MiniMax-H3 DLO component lifecycle (#6526)

2026-08-25 23:13:41 +08:00

test_plan_resolver.py

[Refactor][Diffusion] Resolve offload topology through one plan resolver (#7209)

2026-09-09 08:17:38 +08:00

test_sequential_backend.py

Gate persistent DiT staging on decode-graph models

2026-09-22 08:58:32 +00:00