update

autoregressive/models/generate.py

@@ -138,7 +138,7 @@ def decode_n_tokens(
 
 @torch.no_grad()
 def generate(model, cond, max_new_tokens, emb_masks=None, cfg_scale=1.0, cfg_interval=-1, condition=None, condition_null=None, condition_token_nums=0, **sampling_kwargs):
-    print("cond", torch.any(torch.isnan(cond)))
+    # print("cond", torch.any(torch.isnan(cond)))
     if condition is not None:
         with torch.no_grad():
             # print(f'nan: {torch.any(torch.isnan(model.adapter.model.embeddings.patch_embeddings.projection.weight))}')
@@ -147,7 +147,7 @@ def generate(model, cond, max_new_tokens, emb_masks=None, cfg_scale=1.0, cfg_int
             # print("before condition", condition)
             # condition = torch.ones_like(condition)
             condition = model.adapter_mlp(condition)
-            print("condition", torch.any(torch.isnan(condition)))
+            # print("condition", torch.any(torch.isnan(condition)))
     if model.model_type == 'c2i':
         if cfg_scale > 1.0:
             cond_null = torch.ones_like(cond) * model.num_classes

model.py

@@ -57,7 +57,7 @@ class Model:
 
     def to(self, device):
         self.gpt_model_canny.to('cuda')
-        print(next(self.gpt_model_canny.adapter.parameters()).device)
+        # print(next(self.gpt_model_canny.adapter.parameters()).device)
         # print(self.gpt_model_canny.device)
 
     def load_vq(self):
@@ -88,7 +88,7 @@ class Model:
         # print("prev:", model_weight['adapter.model.embeddings.patch_embeddings.projection.weight'])
         gpt_model.load_state_dict(model_weight, strict=True)
         gpt_model.eval()
-        print("loaded:", gpt_model.adapter.model.embeddings.patch_embeddings.projection.weight)
+        # print("loaded:", gpt_model.adapter.model.embeddings.patch_embeddings.projection.weight)
         print("gpt model is loaded")
         return gpt_model
 
@@ -123,10 +123,11 @@ class Model:
         image = resize_image_to_16_multiple(image, 'canny')
         W, H = image.size
         print(W, H)
+        self.gpt_model_depth.to('cpu')
         self.t5_model.model.to('cuda')
         self.gpt_model_canny.to('cuda')
         self.vq_model.to('cuda')
-        print("after cuda", self.gpt_model_canny.adapter.model.embeddings.patch_embeddings.projection.weight)
+        # print("after cuda", self.gpt_model_canny.adapter.model.embeddings.patch_embeddings.projection.weight)
 
         condition_img = self.get_control_canny(np.array(image), low_threshold,
                                                high_threshold)
@@ -202,6 +203,7 @@ class Model:
         image = resize_image_to_16_multiple(image, 'depth')
         W, H = image.size
         print(W, H)
+        self.gpt_model_canny.to('cpu')
         self.t5_model.model.to(self.device)
         self.gpt_model_depth.to(self.device)
         self.get_control_depth.model.to(self.device)