The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Fix lowvram bug when using a model multiple times in a row.
comfyanonymous3ad3248ad772
diff --git a/comfy/model_management.py b/comfy/model_management.pyindex f6ca252e3..d5d77d7d7 100644--- a/comfy/model_management.py+++ b/comfy/model_management.py@@ -505,7 +505,7 @@ def load_models_gpu(models, memory_required=0, force_patch_weights=False, minimuif lowvram_available and (vram_set_state == VRAMState.LOW_VRAM or vram_set_state == VRAMState.NORMAL_VRAM) and not force_full_load:model_size = loaded_model.model_memory_required(torch_dev)current_free_mem = get_free_memory(torch_dev)- lowvram_model_memory = max(64 * (1024 * 1024), (current_free_mem - minimum_memory_required), min(current_free_mem * 0.4, current_free_mem - minimum_inference_memory()))+ lowvram_model_memory = max(1, (current_free_mem - minimum_memory_required), min(current_free_mem * 0.4, current_free_mem - minimum_inference_memory()))if model_size <= lowvram_model_memory: #only switch to lowvram if really necessarylowvram_model_memory = 0