-
1e076a6523
Update example workflow with attention and denoise parameters
master
Ethanfel
2026-02-15 03:14:58 +01:00
-
91e5bd8222
Clean up debug logging and fix precision setting for autocast
Ethanfel
2026-02-15 03:13:33 +01:00
-
45e57f58a0
Print model load status to detect missing/unexpected weight keys
Ethanfel
2026-02-15 02:34:26 +01:00
-
162c27d243
Add space between user prompt and quality prompt suffix
Ethanfel
2026-02-15 02:32:20 +01:00
-
741c02b88c
Add diagnostic debug logging to pipeline stages
Ethanfel
2026-02-15 02:25:10 +01:00
-
e2025c6ca0
Move VAE encode outside autocast to match original STAR pipeline
Ethanfel
2026-02-15 02:14:25 +01:00
-
0537d9d8a5
Expose denoise parameter (0.1–1.0) in node UI
Ethanfel
2026-02-15 02:06:57 +01:00
-
8a440761d1
Fix noise level (900 not 1000) and prompt concatenation to match original STAR
Ethanfel
2026-02-15 02:03:34 +01:00
-
2bf8db4f07
Use fp32 accumulation in SDPA and math attention to match xformers precision
Ethanfel
2026-02-15 01:47:10 +01:00
-
0508868978
Revert SDPA to 3D tensors — 4D unsqueeze caused quality degradation
Ethanfel
2026-02-15 01:37:10 +01:00
-
f03c4853f1
Revert model loading to original HF-based paths
Ethanfel
2026-02-15 01:20:37 +01:00
-
4c6c38f05a
Fix attention dispatcher: use 4D tensors for SDPA, add math backend
Ethanfel
2026-02-15 01:05:51 +01:00
-
f991f5cb02
Load text encoder and VAE from ComfyUI model folders
Ethanfel
2026-02-15 00:37:14 +01:00
-
e272e1a57d
Fix open_clip batch_first compatibility via auto-applied patch
Ethanfel
2026-02-15 00:25:26 +01:00
-
82d7f4997a
Add configurable attention backend with SageAttention variant support
Ethanfel
2026-02-15 00:12:26 +01:00
-
cf74b587ec
Add SageAttention as preferred attention backend when available
Ethanfel
2026-02-15 00:00:55 +01:00
-
5de26d8ead
Add xformers compatibility shim using PyTorch native SDPA
Ethanfel
2026-02-14 23:58:55 +01:00
-
5786ab6be7
Auto-initialize STAR submodule if missing on first load
Ethanfel
2026-02-14 23:48:43 +01:00
-
6cf314baf4
Add standalone CLI for memory-efficient video upscaling
Ethanfel
2026-02-14 23:46:58 +01:00
-
8794f8ddec
Add example workflow and link in README
Ethanfel
2026-02-14 23:32:35 +01:00
-
f7021e95f4
Add segment-based processing for long videos to reduce RAM usage
Ethanfel
2026-02-14 23:28:01 +01:00
-
5f9287cfac
Initial release: ComfyUI nodes for STAR video super-resolution
Ethanfel
2026-02-14 23:20:27 +01:00