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.

Memory usage factor for ideogram 4 on non dynamic vram. (#14264)

comfyanonymous8e3045a90b4b
diff --git a/comfy/supported_models.py b/comfy/supported_models.py
index 478489ed8..7cf9c133b 100644
--- a/comfy/supported_models.py
+++ b/comfy/supported_models.py
@@ -1757,7 +1757,7 @@ class Ideogram4(supported_models_base.BASE):
"shift": 1.0,
}
- memory_usage_factor = 1.8 # TODO
+ memory_usage_factor = 11.6
unet_extra_config = {
"num_attention_heads": 18,