Cleanup

extensions-builtin/Lora/networks.py

@@ -146,11 +146,15 @@ def load_networks(names, te_multipliers=None, unet_multipliers=None, dyn_dims=No
             for i, name in enumerate(names):
                 if shared.compiled_model_state.lora_model[i] != f"{name}:{te_multipliers[i] if te_multipliers else 1.0}":
                     recompile_model = True
+                    shared.compiled_model_state.lora_model = []
                     break
         else:
             recompile_model = True
-        shared.compiled_model_state.lora_model = []
+            shared.compiled_model_state.lora_model = []
     if recompile_model:
+        if not shared.opts.openvino_disable_model_caching:
+            shared.log.warning("LoRa: Disabling OpenVINO model caching")
+            shared.opts.openvino_disable_model_caching = True
         sd_models.unload_model_weights(op='model')
         shared.opts.cuda_compile = False
         sd_models.reload_model_weights(op='model')

modules/shared.py

@@ -300,7 +300,7 @@ options_templates.update(options_section(('cuda', "Compute Settings"), {
     "ipex_optimize_upscaler": OptionInfo(True if devices.backend == "ipex" else False, "Enable IPEX Optimize for Intel GPUs with Upscalers"),
     "directml_memory_provider": OptionInfo(default_memory_provider, 'DirectML memory stats provider', gr.Radio, {"choices": memory_providers}),
     "directml_catch_nan": OptionInfo(False, "DirectML retry specific operation when NaN is produced if possible. (makes generation slower)"),
-    "openvino_disable_model_caching": OptionInfo(True, "OpenVINO disable model caching"),
+    "openvino_disable_model_caching": OptionInfo(False, "OpenVINO disable model caching"),
     "openvino_hetero_gpu": OptionInfo(False, "OpenVINO use Hetero Device for single inference with multiple devices"),
     "openvino_remove_cpu_from_hetero": OptionInfo(False, "OpenVINO remove CPU from Hetero Device"),
     "openvino_remove_igpu_from_hetero": OptionInfo(False, "OpenVINO remove iGPU from Hetero Device"),