Harden latent upscale GPU cleanup

dumas_h3_longvideos.py

@@ -6055,7 +6055,7 @@ def quant_accel_note(model):
 
 
 
-def _evict_all_but(keep_model):
+def _evict_all_but(keep_model):
     """Unload every model EXCEPT the diffusion model from the GPU.
 
     This is the fix for VRAM ratcheting across a long chain. soft_empty_cache()
@@ -6081,15 +6081,38 @@ def _evict_all_but(keep_model):
     except Exception:
         try:
             mm.soft_empty_cache(True)
-        except Exception:
-            pass
-
-
-
-
-
-
-class H3LongVideos:
+        except Exception:
+            pass
+
+
+def _evict_for_latent_upscale(model):
+    """Clear the sampler model before loading the auxiliary latent upscaler."""
+    try:
+        unload_clones = getattr(mm, "unload_model_and_clones", None)
+        if callable(unload_clones):
+            try:
+                unload_clones(model, unload_additional_models=False)
+                mm.soft_empty_cache()
+                return
+            except Exception:
+                pass
+        mm.unload_all_models()
+    except Exception:
+        pass
+    try:
+        mm.soft_empty_cache(True)
+    except Exception:
+        try:
+            mm.soft_empty_cache()
+        except Exception:
+            pass
+
+
+
+
+
+
+class H3LongVideos:
     CATEGORY = "Dumas/MiniMax"
     FUNCTION = "run"
     # fps is emitted as BOTH types on purpose: ComfyUI does not coerce between them,
@@ -6662,9 +6685,8 @@ class H3LongVideos:
                 parts = _nested_tensor_parts(out_samples)
                 if not getattr(out_samples, "is_nested", False) or len(parts) < 2:
                     raise RuntimeError("latent upscale expects a nested AV latent")
-                if latent_upscale_mode == "model" and str(latent_upscale_param.get("device", "cuda")) == "cuda" and hasattr(model, "clone_base_uuid"):
-                    mm.unload_model_and_clones(model, unload_additional_models=False)
-                    mm.soft_empty_cache()
+                if latent_upscale_mode == "model" and str(latent_upscale_param.get("device", "cuda")) == "cuda":
+                    _evict_for_latent_upscale(model)
                 upscaled_video, up_h, up_w = _upscale_latent_video(parts[0], latent_upscale_param)
                 full_audio = parts[1]
                 # Drop the first-pass sampling state before we start the refinement