Update app.py

app.py

@@ -158,19 +158,7 @@ def upload_to_ftp(filename):
     except Exception as e:
         print(f"FTP upload error: {e}")
 
-def apply_style(style_name: str, positive: str, negative: str = "") -> Tuple[str, str]:
-    if style_name in styles:
-        p, n = styles.get(style_name, styles[DEFAULT_STYLE_NAME])
-    else:
-        p, n = styles[DEFAULT_STYLE_NAME]
-    if not negative:
-        negative = ""
-    return p.replace("{prompt}", positive), n + negative
-    
-def save_image(img):
-    unique_name = str(uuid.uuid4()) + ".png"
-    img.save(unique_name,optimize=False,compress_level=0)
-    return unique_name
+
 
 def uploadNote(prompt,num_inference_steps,guidance_scale,timestamp):
     filename= f'rv_C_{timestamp}.txt'
@@ -187,6 +175,56 @@ def uploadNote(prompt,num_inference_steps,guidance_scale,timestamp):
     upload_to_ftp(filename) 
 
 code = r'''
+
+def scheduler_swap_callback(pipeline, step_index, timestep, callback_kwargs):
+        # adjust the batch_size of prompt_embeds according to guidance_scale
+    if step_index == int(pipeline.num_timesteps * 0.1):
+        print("-- swapping scheduler --")
+      #  pipeline.scheduler = euler_scheduler
+        torch.set_float32_matmul_precision("high")
+       # pipe.vae = vae_b
+        torch.backends.cudnn.allow_tf32 = True
+        torch.backends.cuda.matmul.allow_tf32 = True
+        torch.backends.cudnn.deterministic = True
+        torch.backends.cuda.preferred_blas_library="cublaslt"
+    #if step_index == int(pipeline.num_timesteps * 0.5):
+       # torch.set_float32_matmul_precision("medium")
+        #callback_kwargs["latents"] = callback_kwargs["latents"].to(torch.float64)
+        #pipe.unet.to(torch.float64)
+      #  pipe.guidance_scale=1.0
+       # pipe.scheduler.set_timesteps(num_inference_steps*.70)
+      #  print(f"-- setting step {pipeline.num_timesteps * 0.1} --")
+      #  pipeline.scheduler._step_index = pipeline.num_timesteps * 0.1 
+    if step_index == int(pipeline.num_timesteps * 0.9):
+        torch.backends.cuda.preferred_blas_library="cublas"
+        torch.backends.cudnn.allow_tf32 = False
+        torch.backends.cuda.matmul.allow_tf32 = False
+        torch.set_float32_matmul_precision("highest")
+        #callback_kwargs["latents"] = callback_kwargs["latents"].to(torch.bfloat16)
+        #pipe.unet.to(torch.float64)
+     #   pipe.vae = vae_a
+     #   pipe.unet = unet_a
+        torch.backends.cudnn.deterministic = False
+        print("-- swapping scheduler --")
+      #  pipeline.scheduler = heun_scheduler
+        #pipe.scheduler.set_timesteps(num_inference_steps*.70)
+      #  print(f"-- setting step {pipeline.num_timesteps * 0.9} --")
+      #  pipeline.scheduler._step_index = pipeline.num_timesteps * 0.9
+    return {"latents": callback_kwargs["latents"]}   
+
+def upload_to_ftp(filename):
+    try:
+        transport = paramiko.Transport((FTP_HOST, 22))
+        destination_path=FTP_DIR+filename
+        transport.connect(username = FTP_USER, password = FTP_PASS)
+        sftp = paramiko.SFTPClient.from_transport(transport)
+        sftp.put(filename, destination_path)
+        sftp.close()
+        transport.close()
+        print(f"Uploaded {filename} to FTP server")
+    except Exception as e:
+        print(f"FTP upload error: {e}")
+        
 def uploadNote(prompt,num_inference_steps,guidance_scale,timestamp):
     filename= f'rv_C_{timestamp}.txt'
     with open(filename, "w") as f:
@@ -200,6 +238,7 @@ def uploadNote(prompt,num_inference_steps,guidance_scale,timestamp):
         f.write(f"Model VAE: sdxl-vae-bf16\n")
         f.write(f"To cuda and bfloat \n")
     return filename
+    
 '''
 
 pyx = cyper.inline(code)
@@ -229,20 +268,21 @@ def generate_30(
         "num_inference_steps": num_inference_steps,
         "generator": generator,
         "output_type": "pil",
-        "callback_on_step_end": scheduler_swap_callback
+        "callback_on_step_end": pyx.scheduler_swap_callback
     }
     if use_resolution_binning:
         options["use_resolution_binning"] = True
     images = []
     timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
     filename = pyx.uploadNote(prompt,num_inference_steps,guidance_scale,timestamp)
-    upload_to_ftp(filename) 
+    #upload_to_ftp(filename) 
+    pyx.upload_to_ftp(filename) 
     #uploadNote(prompt,num_inference_steps,guidance_scale,timestamp)
     batch_options = options.copy()
     rv_image = pipe(**batch_options).images[0]
     sd_image_path = f"rv_C_{timestamp}.png"
     rv_image.save(sd_image_path,optimize=False,compress_level=0)
-    upload_to_ftp(sd_image_path)    
+    pyx.upload_to_ftp(sd_image_path)    
     unique_name = str(uuid.uuid4()) + ".png"  
     os.symlink(sd_image_path, unique_name)  
     return [unique_name]