PozzettiAndrea
3cc4d96d65
updoot
2026-04-16 01:14:58 +02:00
PozzettiAndrea
b768bd9836
att
2026-02-21 16:18:12 +01:00
PozzettiAndrea
8f1be7335a
add memory dropdown (cache_gpu, cpu_offload, delete) to LoadSAM3DBodyModel
...
Adds model memory management with three strategies: keep on GPU, offload
to CPU RAM, or delete after inference. Updates all workflow JSONs.
2026-02-15 18:12:52 +01:00
PozzettiAndrea
4ff3fbe409
remove unused server routes that caused CUDA crash on CPU runners
...
The server.py endpoints (/sam3d/save_glb, /sam3d/export_posed_fbx) were
never called by the frontend. Importing PromptServer at module level
triggered model_management → torch.cuda.current_device() → access
violation on CPU-only machines during instantiation tests.
2026-02-15 16:36:20 +01:00
PozzettiAndrea
5acc182dd3
use flash attention
2026-02-15 15:58:45 +01:00
Andrea Pozzetti
ac2fc601c2
partial progress
2025-12-28 15:35:55 +01:00
Andrea Pozzetti
abcd5710df
added loaders for fbx
2025-11-23 05:59:38 +01:00
Andrea Pozzetti
36cf8db318
added rigging
2025-11-23 05:58:38 +01:00
PozzettiAndrea