refine model_unload (#14)

Refine model_unload behavior to improve robustness and offload precision.

- Normalize memory_to_free before logging to avoid None / invalid arithmetic
- Make partially_unload more precise by capping memory_to_free to loaded size
- Ensure full unload cases are handled correctly after partial unload

Author: sfiisf

comfy/model_management.py

@@ -546,21 +546,22 @@ def should_reload_model(self, force_patch_weights=False):
         return False
 
     def model_unload(self, memory_to_free=None, unpatch_weights=True):
+        model_loaded_size = self.model.loaded_size()
+        if memory_to_free is None:
+            # free the full model
+            memory_to_free = model_loaded_size
+
         logging.debug(f"model_unload: {self.model.model.__class__.__name__}")
         logging.debug(f"memory_to_free: {memory_to_free/(1024*1024*1024)} GB")
         logging.debug(f"unpatch_weights: {unpatch_weights}")
-        logging.debug(f"loaded_size: {self.model.loaded_size()/(1024*1024*1024)} GB")
+        logging.debug(f"loaded_size: {model_loaded_size/(1024*1024*1024)} GB")
         logging.debug(f"offload_device: {self.model.offload_device}")
 
-        if memory_to_free is None:
-            # free the full model
-            memory_to_free = self.model.loaded_size()
-
         available_memory = get_free_memory(self.model.offload_device)
         logging.debug(f"before unload, available_memory of offload device {self.model.offload_device}: {available_memory/(1024*1024*1024)} GB")
 
         mmap_mem_threshold = get_mmap_mem_threshold_gb() * 1024 * 1024 * 1024  # this is reserved memory for other system usage
-        if memory_to_free > available_memory - mmap_mem_threshold or memory_to_free < self.model.loaded_size():
+        if min(memory_to_free, model_loaded_size) > available_memory - mmap_mem_threshold or memory_to_free < model_loaded_size:
             partially_unload = True
         else:
             partially_unload = False
@@ -571,6 +572,8 @@ def model_unload(self, memory_to_free=None, unpatch_weights=True):
             logging.debug(f"partially_unload freed vram: {freed/(1024*1024*1024)} GB")
             if freed < memory_to_free:
                 logging.warning(f"Partially unload not enough memory, freed {freed/(1024*1024*1024)} GB, memory_to_free {memory_to_free/(1024*1024*1024)} GB")
+            if freed == model_loaded_size:
+                partially_unload = False
         else:
             logging.debug("Do full unload")
             self.model.detach(unpatch_weights)