Make sure process function accesses model

model.py

@@ -76,15 +76,15 @@ class Model:
         subprocess.run(shlex.split(f'wget {pidinet_file} -O models/table5_pidinet.pth'))
 
         
-        model = load_model_from_config(config, "models/sd-v1-4.ckpt").to(device)
+        self.model = load_model_from_config(config, "models/sd-v1-4.ckpt").to(device)
         current_base = 'sd-v1-4.ckpt'
-        model_ad = Adapter(channels=[320, 640, 1280, 1280][:4], nums_rb=2, ksize=1, sk=True, use_conv=False).to(device)
-        model_ad.load_state_dict(torch.load("models/t2iadapter_sketch_sd14v1.pth"))
+        self.model_ad = Adapter(channels=[320, 640, 1280, 1280][:4], nums_rb=2, ksize=1, sk=True, use_conv=False).to(device)
+        self.model_ad.load_state_dict(torch.load("models/t2iadapter_sketch_sd14v1.pth"))
         net_G = pidinet()
         ckp = torch.load('models/table5_pidinet.pth', map_location='cpu')['state_dict']
         net_G.load_state_dict({k.replace('module.',''):v for k, v in ckp.items()})
         net_G.to(device)
-        sampler = PLMSSampler(model)
+        self.sampler = PLMSSampler(self.model)
         save_memory=True
 
     @torch.inference_mode()
@@ -121,17 +121,17 @@ class Model:
             im = im.float()
             im_edge = tensor2img(im)
     
-        c = model.get_learned_conditioning([prompt])
-        nc = model.get_learned_conditioning([neg_prompt])
+        c = self.model.get_learned_conditioning([prompt])
+        nc = self.model.get_learned_conditioning([neg_prompt])
         
         with torch.no_grad():
             # extract condition features
-            features_adapter = model_ad(im.to(device))
+            features_adapter = self.model_ad(im.to(device))
     
         shape = [4, 64, 64]
     
         # sampling
-        samples_ddim, _ = sampler.sample(S=50,
+        samples_ddim, _ = self.sampler.sample(S=50,
                                         conditioning=c,
                                         batch_size=1,
                                         shape=shape,
@@ -144,7 +144,7 @@ class Model:
                                         mode = 'sketch',
                                         con_strength = con_strength)
     
-        x_samples_ddim = model.decode_first_stage(samples_ddim)
+        x_samples_ddim = self.model.decode_first_stage(samples_ddim)
         x_samples_ddim = torch.clamp((x_samples_ddim + 1.0) / 2.0, min=0.0, max=1.0)
         x_samples_ddim = x_samples_ddim.permute(0, 2, 3, 1).numpy()[0]
         x_samples_ddim = 255.*x_samples_ddim