Better reporting of model load errors - sends the report to the browser UI during the next image rendering task

ui/easydiffusion/model_manager.py

@@ -53,15 +53,21 @@ def load_default_models(context: Context):
                 scan_model=context.model_paths[model_type] != None
                 and not context.model_paths[model_type].endswith(".safetensors"),
             )
+            if model_type in context.model_load_errors:
+                del context.model_load_errors[model_type]
         except Exception as e:
             log.error(f"[red]Error while loading {model_type} model: {context.model_paths[model_type]}[/red]")
             log.exception(e)
             del context.model_paths[model_type]
 
+            context.model_load_errors[model_type] = str(e)  # storing the entire Exception can lead to memory leaks
+
 
 def unload_all(context: Context):
     for model_type in KNOWN_MODEL_TYPES:
         unload_model(context, model_type)
+        if model_type in context.model_load_errors:
+            del context.model_load_errors[model_type]
 
 
 def resolve_model_to_use(model_name: str = None, model_type: str = None):
@@ -132,7 +138,14 @@ def reload_models_if_necessary(context: Context, task_data: TaskData):
         context.model_paths[model_type] = model_path_in_req
 
         action_fn = unload_model if context.model_paths[model_type] is None else load_model
+        try:
             action_fn(context, model_type, scan_model=False)  # we've scanned them already
+            if model_type in context.model_load_errors:
+                del context.model_load_errors[model_type]
+        except Exception as e:
+            log.exception(e)
+            if action_fn == load_model:
+                context.model_load_errors[model_type] = str(e)  # storing the entire Exception can lead to memory leaks
 
 
 def resolve_model_paths(task_data: TaskData):
@@ -149,6 +162,14 @@ def resolve_model_paths(task_data: TaskData):
         task_data.use_upscale = resolve_model_to_use(task_data.use_upscale, "realesrgan")
 
 
+def fail_if_models_did_not_load(context: Context):
+    for model_type in KNOWN_MODEL_TYPES:
+        if model_type in context.model_load_errors:
+            e = context.model_load_errors[model_type]
+            raise Exception(f"Could not load the {model_type} model! Reason: " + e)
+            # concat 'e', don't use in format string (injection attack)
+
+
 def set_vram_optimizations(context: Context):
     config = app.getConfig()
     vram_usage_level = config.get("vram_usage_level", "balanced")

ui/easydiffusion/renderer.py

@@ -33,6 +33,7 @@ def init(device):
     context.stop_processing = False
     context.temp_images = {}
     context.partial_x_samples = None
+    context.model_load_errors = {}
 
     from easydiffusion import app
 

ui/easydiffusion/task_manager.py

@@ -336,6 +336,7 @@ def thread_render(device):
             current_state = ServerStates.LoadingModel
             model_manager.resolve_model_paths(task.task_data)
             model_manager.reload_models_if_necessary(renderer.context, task.task_data)
+            model_manager.fail_if_models_did_not_load(renderer.context)
 
             current_state = ServerStates.Rendering
             task.response = renderer.make_images(