[BE] supports to customize gpu device name when bootstrapping backend

backend.py

@@ -40,7 +40,9 @@ logger = Logger(name=LOGGER_NAME_BACKEND)
 database = Database(logger)
 
 
-def load_model(logger: Logger, use_gpu: bool, reduce_memory_usage: bool) -> Model:
+def load_model(
+    logger: Logger, use_gpu: bool, gpu_device_name: str, reduce_memory_usage: bool
+) -> Model:
     # model candidates:
     # "runwayml/stable-diffusion-v1-5"
     # "CompVis/stable-diffusion-v1-4"
@@ -56,7 +58,13 @@ def load_model(logger: Logger, use_gpu: bool, reduce_memory_usage: bool) -> Mode
     # "runwayml/stable-diffusion-inpainting"
     inpainting_model_name = "runwayml/stable-diffusion-inpainting"
 
-    model = Model(model_name, inpainting_model_name, logger, use_gpu=use_gpu)
+    model = Model(
+        model_name,
+        inpainting_model_name,
+        logger,
+        use_gpu=use_gpu,
+        gpu_device_name=gpu_device_name,
+    )
     if use_gpu and reduce_memory_usage:
         model.set_low_memory_mode()
     model.load_all()
@@ -172,7 +180,7 @@ def main(args):
     database.set_image_output_folder(args.image_output_folder)
     database.connect(args.db)
 
-    model = load_model(logger, args.gpu, args.reduce_memory_usage)
+    model = load_model(logger, args.gpu, args.gpu_device, args.reduce_memory_usage)
     backend(model, args.gfpgan, args.debug)
 
     database.safe_disconnect()
@@ -189,9 +197,14 @@ if __name__ == "__main__":
         "--db", type=str, default="happysd.db", help="Path to SQLite database file"
     )
 
-    # Add an argument to set the path of the database file
+    # Add an argument to set the 'gpu' flag
     parser.add_argument("--gpu", action="store_true", help="Enable to use GPU device")
 
+    # Add an argument to set the gpu device name
+    parser.add_argument(
+        "--gpu-device", type=str, default="cuda", help="GPU device name"
+    )
+
     # Add an argument to reduce memory usage
     parser.add_argument(
         "--reduce-memory-usage",

utilities/img2img.py

@@ -31,7 +31,7 @@ class Img2Img:
         logger: DummyLogger = DummyLogger(),
     ):
         self.model = model
-        self.__device = "cpu" if not self.model.use_gpu() else "cuda"
+        self.__device = "cpu" if not self.model.use_gpu() else self.model.get_gpu_device_name()
         self.__output_folder = output_folder
         self.__logger = logger
 

utilities/inpainting.py

@@ -31,7 +31,7 @@ class Inpainting:
         logger: DummyLogger = DummyLogger(),
     ):
         self.model = model
-        self.__device = "cpu" if not self.model.use_gpu() else "cuda"
+        self.__device = "cpu" if not self.model.use_gpu() else self.model.get_gpu_device_name()
         self.__output_folder = output_folder
         self.__logger = logger
 

utilities/model.py

@@ -24,13 +24,15 @@ class Model:
         inpainting_model_name: str,
         logger: DummyLogger = DummyLogger(),
         use_gpu: bool = True,
+        gpu_device_name: str = "cuda",
     ):
         self.model_name = model_name
         self.inpainting_model_name = inpainting_model_name
         self.__use_gpu = False
+        self.__gpu_device = gpu_device_name
         if use_gpu and torch.cuda.is_available():
             self.__use_gpu = True
-            logger.info("running on {}".format(torch.cuda.get_device_name("cuda:0")))
+            logger.info("running on {}".format(torch.cuda.get_device_name(self.__gpu_device)))
         else:
             logger.info("running on CPU (expect it to be verrry sloooow)")
         self.__logger = logger
@@ -44,6 +46,9 @@ class Model:
     def use_gpu(self):
         return self.__use_gpu
     
+    def get_gpu_device_name(self):
+        return self.__gpu_device
+    
     def update_model_name(self, model_name:str):
         if not model_name or model_name == self.model_name:
             self.__logger.warn("model name empty or the same, not updated")
@@ -113,7 +118,7 @@ class Model:
                     "failed to load model %s: %s" % (self.model_name, e)
                 )
         if pipeline and self.use_gpu():
-            pipeline.to("cuda")
+            pipeline.to(self.get_gpu_device_name())
             
         self.txt2img_pipeline = pipeline
         self.__default_txt2img_scheduler = pipeline.scheduler
@@ -154,7 +159,7 @@ class Model:
                     % (self.inpainting_model_name, e)
                 )
         if pipeline and self.use_gpu():
-            pipeline.to("cuda")
+            pipeline.to(self.get_gpu_device_name())
         self.inpaint_pipeline = pipeline
         self.__default_inpaint_scheduler = pipeline.scheduler
         empty_memory_cache()

utilities/text2img.py

@@ -28,7 +28,7 @@ class Text2Img:
         logger: DummyLogger = DummyLogger(),
     ):
         self.model = model
-        self.__device = "cpu" if not self.model.use_gpu() else "cuda"
+        self.__device = "cpu" if not self.model.use_gpu() else self.model.get_gpu_device_name()
         self.__output_folder = output_folder
         self.__logger = logger