Update app.py

app.py

@@ -10,7 +10,7 @@ import matplotlib.pyplot as plt
 import numpy as np
 import PIL
 
-torch.cuda.empty_cache()
+
 
 model = whisper.load_model("base")
 model.device
@@ -18,7 +18,7 @@ model.device
 model_id = 'prompthero/midjourney-v4-diffusion' #"stabilityai/stable-diffusion-2"
 
 scheduler = EulerDiscreteScheduler.from_pretrained(model_id, subfolder="scheduler")
-pipe = StableDiffusionPipeline.from_pretrained(model_id) #pipe = StableDiffusionPipeline.from_pretrained(model_id , torch_dtype=torch.float16 #pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, revision="fp16", torch_dtype=torch.float16)
+pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.torch.cuda.empty_cache()16) #pipe = StableDiffusionPipeline.from_pretrained(model_id , torch_dtype=torch.float16 #pipe = StableDiffusionPipeline.from_pretrained(model_id, scheduler=scheduler, revision="fp16", torch_dtype=torch.float16)
 pipe = pipe.to("cuda")
 
 def transcribe(audio,prompt_num,user_keywords):
@@ -175,11 +175,12 @@ def keywords(text,prompt_num,user_keywords):
             count += 1
             print(i)
             print("works4")
-            # with torch.autocast("cuda"):
-            image = pipe(i, height=768, width=768, guidance_scale = 10).images[0]
-            print("works5")
-            images.append(image)
-            print("works6")
+            torch.cuda.empty_cache()
+            with torch.autocast("cuda"):
+                image = pipe(i, height=768, width=768, guidance_scale = 10).images[0]
+                print("works5")
+                images.append(image)
+                print("works6")
     
     # min_shape = sorted( [(np.sum(i.size), i.size ) for i in images])[0][1]
     # imgs_comb = np.hstack([i.resize(min_shape) for i in images])