Ethanfel/Comfyui-STAR
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Clean up debug logging and fix precision setting for autocast

Browse Source 
Remove all [STAR DEBUG] print statements added during quality
investigation. Fix autocast to actually use the selected precision
dtype (fp16/bf16) instead of always defaulting to fp16. fp32 now
properly disables autocast for full-precision inference.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Ethanfel
2026-02-15 03:13:33 +01:00
parent 45e57f58a0
commit 91e5bd8222
2 changed files with 4 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
nodes.py
View File

@@ -186,7 +186,7 @@ class STARModelLoader:
}),
"precision": (["fp16", "bf16", "fp32"], {
"default": "fp16",
"tooltip": "Weight precision. fp16 is recommended (fastest, lowest VRAM). bf16 for newer GPUs. fp32 for maximum quality at 2x VRAM cost.",
"tooltip": "Inference precision. fp16 is recommended (fastest, lowest VRAM). bf16 for newer GPUs. fp32 disables autocast for maximum quality at 2x VRAM cost.",
}),
"offload": (["disabled", "model", "aggressive"], {
"default": "disabled",