Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Increase trellis2 memory factor a bit. (#15796)

comfyanonymous7dde56176efa
diff --git a/comfy/supported_models.py b/comfy/supported_models.py
index 1dc06c9cc..d2b4e6af2 100644
--- a/comfy/supported_models.py
+++ b/comfy/supported_models.py
@@ -1489,7 +1489,7 @@ class Trellis2(supported_models_base.BASE):
"shift": 3.0,
}
- memory_usage_factor = 3.5
+ memory_usage_factor = 6
latent_format = latent_formats.Trellis2
vae_key_prefix = ["vae."]