diff --git a/core/leras/nn.py b/core/leras/nn.py
index 280bcdb..09900bf 100644
--- a/core/leras/nn.py
+++ b/core/leras/nn.py
@@ -76,15 +76,19 @@ class nn():
             
             first_run = False
             
-            if sys.platform[0:3] == 'win':
-                devices_str = ""
-                for device in device_config.devices:
-                    devices_str += "_" + device.name.replace(' ','_')
-                
-                compute_cache_path = Path(os.environ['APPDATA']) / 'NVIDIA' / ('ComputeCache' + devices_str)
-                if not compute_cache_path.exists():
-                    first_run = True
-                os.environ['CUDA_CACHE_PATH'] = str(compute_cache_path)
+            if not device_config.cpu_only:
+                if sys.platform[0:3] == 'win':
+                    if all( [ x.name == device_config.devices[0].name for x in device_config.devices ] ):
+                        devices_str = "_" + device_config.devices[0].name.replace(' ','_')
+                    else:
+                        devices_str = ""
+                        for device in device_config.devices:
+                            devices_str += "_" + device.name.replace(' ','_')
+                        
+                    compute_cache_path = Path(os.environ['APPDATA']) / 'NVIDIA' / ('ComputeCache' + devices_str)
+                    if not compute_cache_path.exists():
+                        first_run = True
+                    os.environ['CUDA_CACHE_PATH'] = str(compute_cache_path)
 
             os.environ['TF_MIN_GPU_MULTIPROCESSOR_COUNT'] = '2'
             os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' # tf log errors only
diff --git a/mainscripts/Extractor.py b/mainscripts/Extractor.py
index 7d41684..7b40f2a 100644
--- a/mainscripts/Extractor.py
+++ b/mainscripts/Extractor.py
@@ -156,6 +156,8 @@ class ExtractSubprocessor(Subprocessor):
                             landmarks_extractor,
                             rects_extractor,
                             ):
+            h, w, ch = image.shape
+            
             if data.rects_rotation == 0:
                 rotated_image = image
             elif data.rects_rotation == 90:
diff --git a/models/Model_SAEHD/Model.py b/models/Model_SAEHD/Model.py
index 96e553d..b676a61 100644
--- a/models/Model_SAEHD/Model.py
+++ b/models/Model_SAEHD/Model.py
@@ -615,9 +615,9 @@ class SAEHDModel(ModelBase):
             with tf.device( f'/GPU:0' if len(devices) != 0 else f'/CPU:0'):
                 if 'df' in archi:                
                     gpu_dst_code     = self.inter(self.encoder(self.warped_dst))
-                    gpu_pred_src_dst = self.decoder_src(gpu_dst_code)
-                    gpu_pred_dst_dstm = self.decoder_dstm(gpu_dst_code)
-                    gpu_pred_src_dstm = self.decoder_srcm(gpu_dst_code)
+                    gpu_pred_src_dst, gpu_pred_src_dstm = self.decoder_src(gpu_dst_code)
+                    _, gpu_pred_dst_dstm = self.decoder_dst(gpu_dst_code)
+                    
                 elif 'liae' in archi:
                     gpu_dst_code = self.encoder (self.warped_dst)
                     gpu_dst_inter_B_code = self.inter_B (gpu_dst_code)
@@ -625,9 +625,8 @@ class SAEHDModel(ModelBase):
                     gpu_dst_code = tf.concat([gpu_dst_inter_B_code,gpu_dst_inter_AB_code],-1)
                     gpu_src_dst_code = tf.concat([gpu_dst_inter_AB_code,gpu_dst_inter_AB_code],-1)
 
-                    gpu_pred_src_dst = self.decoder(gpu_src_dst_code)
-                    gpu_pred_dst_dstm = self.decoderm(gpu_dst_code)
-                    gpu_pred_src_dstm = self.decoderm(gpu_src_dst_code)
+                    gpu_pred_src_dst, gpu_pred_src_dstm = self.decoder(gpu_src_dst_code)
+                    _, gpu_pred_dst_dstm = self.decoder(gpu_dst_code)
                     
             if learn_mask:
                 def AE_merge( warped_dst):