diff --git a/automatic.sh b/automatic.sh
index 9ae467a08..9e8d0341e 100755
--- a/automatic.sh
+++ b/automatic.sh
@@ -1,4 +1,4 @@
 #/bin/env bash
 
 export PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512
-python launch.py --api --xformers --disable-console-progressbars
+python launch.py --api --xformers --disable-console-progressbars --load_gpu
diff --git a/modules/sd_models.py b/modules/sd_models.py
index 76a89e88c..87ef1e652 100644
--- a/modules/sd_models.py
+++ b/modules/sd_models.py
@@ -170,13 +170,13 @@ def get_state_dict_from_checkpoint(pl_sd):
 
 def read_state_dict(checkpoint_file, print_global_state=False, map_location=None):
     _, extension = os.path.splitext(checkpoint_file)
+    device = map_location or shared.weight_load_location
+    if device is None:
+        device = devices.get_cuda_device_string() if torch.cuda.is_available() else "cpu"
     if extension.lower() == ".safetensors":
-        device = map_location or shared.weight_load_location
-        if device is None:
-            device = devices.get_cuda_device_string() if torch.cuda.is_available() else "cpu"
         pl_sd = safetensors.torch.load_file(checkpoint_file, device=device)
     else:
-        pl_sd = torch.load(checkpoint_file, map_location=map_location or shared.weight_load_location)
+        pl_sd = torch.load(checkpoint_file, map_location=device)
 
     if print_global_state and "global_step" in pl_sd:
         print(f"Global Step: {pl_sd['global_step']}")
diff --git a/modules/shared.py b/modules/shared.py
index a6712dae9..c00c10efa 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -89,6 +89,7 @@ parser.add_argument("--api-auth", type=str, help='Set authentication for API lik
 parser.add_argument("--api-log", action='store_true', help="use api-log=True to enable logging of all API requests")
 parser.add_argument("--nowebui", action='store_true', help="use api=True to launch the API instead of the webui")
 parser.add_argument("--ui-debug-mode", action='store_true', help="Don't load model to quickly launch UI")
+parser.add_argument("--load_gpu",  action='store_true', help="Force loading model in GPU memory")
 parser.add_argument("--device-id", type=str, help="Select the default CUDA device to use (export CUDA_VISIBLE_DEVICES=0,1,etc might be needed before)", default=None)
 parser.add_argument("--administrator", action='store_true', help="Administrator rights", default=False)
 parser.add_argument("--cors-allow-origins", type=str, help="Allowed CORS origin(s) in the form of a comma-separated list (no spaces)", default=None)
@@ -131,7 +132,12 @@ devices.device, devices.device_interrogate, devices.device_gfpgan, devices.devic
     (devices.cpu if any(y in cmd_opts.use_cpu for y in [x, 'all']) else devices.get_optimal_device() for x in ['sd', 'interrogate', 'gfpgan', 'esrgan', 'codeformer'])
 
 device = devices.device
-weight_load_location = None if cmd_opts.lowram else "cpu"
+if cmd_opts.lowram:
+    weight_load_location = None
+elif cmd_opts.load_gpu:
+    weight_load_location = device
+else:
+    weight_load_location = "cpu"
 
 batch_cond_uncond = cmd_opts.always_batch_cond_uncond or not (cmd_opts.lowvram or cmd_opts.medvram)
 parallel_processing_allowed = not cmd_opts.lowvram and not cmd_opts.medvram