Hot FixL: Revert aggresive memory management until a more targeted approach can be developed. This was causing OOMs on models that should load normally using the normal loader.

Update version to 2.4.4.

Author: pollockjj

distorch_2.py

@@ -67,15 +67,12 @@ def new_partially_load(self, device_to, extra_memory=0, full_load=False, force_p
             allocations = safetensor_allocation_store.get(debug_hash)
 
             if not hasattr(self.model, '_distorch_high_precision_loras') or not allocations:
-                mm.unload_all_models()
-                soft_empty_cache_multigpu(logger)
                 result = original_partially_load(self, device_to, extra_memory, force_patch_weights)
                 if hasattr(self, '_distorch_block_assignments'):
                     del self._distorch_block_assignments
                 return result
 
-            mm.unload_all_models()
-            soft_empty_cache_multigpu(logger)
+            # soft_empty_cache_multigpu(logger)
 
             mem_counter = 0
 

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "comfyui-multigpu"
 description = "Provides a suite of custom nodes to manage multiple GPUs for ComfyUI, including advanced model offloading for both GGUF and Safetensor formats with DisTorch, and bespoke MultiGPU support for WanVideoWrapper and other custom nodes."
-version = "2.4.3"
+version = "2.4.4"
 license = {file = "LICENSE"}
 
 [project.urls]