Eric Rollei
963d09c3bf
- Remove non-functional keep_in_cache parameter from all loader nodes - Simplify HunyuanModelCache.store() to always cache models - Add trigger output to generation nodes (HunyuanImage3Generate, HunyuanImage3GenerateLarge) - Add optional trigger input to HunyuanImage3Unload node for workflow sequencing - Enable proper control over when models are cleared from VRAM This allows users to connect generation nodes to unload nodes via the trigger output/input, ensuring cache clearing happens only after image generation completes.
17 lines
219 B
Plaintext
17 lines
219 B
Plaintext
{
|
|
"folders": [
|
|
{
|
|
"path": "."
|
|
},
|
|
{
|
|
"path": "../../../../hunyuan_quantization"
|
|
},
|
|
{
|
|
"path": "../../models/HunyuanImage-3"
|
|
},
|
|
{
|
|
"path": "../../models/HunyuanImage-3-NF4"
|
|
}
|
|
],
|
|
"settings": {}
|
|
} |