Update app.py

app.py

@@ -103,13 +103,9 @@ def apply_style(style_name: str, positive: str, negative: str = "") -> Tuple[str
 def load_and_prepare_model():
     #vae = AutoencoderKL.from_pretrained("ford442/sdxl-vae-bf16", safety_checker=None)
     vaeX = AutoencoderKL.from_pretrained("stabilityai/sdxl-vae", safety_checker=None,use_safetensors=False)
-    #vae = AutoencoderKL.from_pretrained('cross-attention/asymmetric-autoencoder-kl-x-2',use_safetensors=False)
     #vae = AutoencoderKL.from_single_file('https://huggingface.co/ford442/sdxl-vae-bf16/mySLR/myslrVAE_v10.safetensors')
-    #vaeX = AutoencoderKL.from_pretrained("stabilityai/sd-vae-ft-mse",use_safetensors=True)
     #vaeX = AutoencoderKL.from_pretrained('ford442/Juggernaut-XI-v11-fp32',subfolder='vae')   #  ,use_safetensors=True  FAILS
-    #vaeX = AutoencoderKL.from_pretrained('ford442/RealVisXL_V5.0_FP64',subfolder='vae').to(torch.bfloat16)   #  ,use_safetensors=True  FAILS
     #unetX = UNet2DConditionModel.from_pretrained('ford442/RealVisXL_V5.0_BF16',subfolder='unet').to(torch.bfloat16)   #  ,use_safetensors=True  FAILS
-  #  vae = AutoencoderKL.from_pretrained("BeastHF/MyBack_SDXL_Juggernaut_XL_VAE/MyBack_SDXL_Juggernaut_XL_VAE_V10(version_X).safetensors",safety_checker=None).to(torch.bfloat16)
     #sched = EulerAncestralDiscreteScheduler.from_pretrained("SG161222/RealVisXL_V5.0", subfolder='scheduler',beta_schedule="scaled_linear", steps_offset=1,timestep_spacing="trailing"))
     #sched = EulerAncestralDiscreteScheduler.from_pretrained("SG161222/RealVisXL_V5.0", subfolder='scheduler', steps_offset=1,timestep_spacing="trailing")
     sched = EulerAncestralDiscreteScheduler.from_pretrained('ford442/RealVisXL_V5.0_BF16', subfolder='scheduler',beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset=1,use_karras_sigmas=True)
@@ -145,8 +141,7 @@ def load_and_prepare_model():
     #pipe.vae = AutoencoderKL.from_pretrained('stabilityai/sdxl-vae',subfolder='vae',force_upcast=False,scaling_factor= 0.182158767676)  
     #pipe.vae.to(torch.bfloat16)
 
-
-    
+   
     '''
     scaling_factor (`float`, *optional*, defaults to 0.18215):
             The component-wise standard deviation of the trained latent space computed using the first batch of the
@@ -162,7 +157,6 @@ def load_and_prepare_model():
         
     '''
     
-    #sched = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear",use_karras_sigmas=True, algorithm_type="dpmsolver++")
     #pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset=1)
     #pipe.scheduler = DPMSolverMultistepScheduler.from_pretrained('SG161222/RealVisXL_V5.0', subfolder='scheduler', algorithm_type='sde-dpmsolver++')
     pipe.vae = vaeX.to(torch.bfloat16)
@@ -171,18 +165,15 @@ def load_and_prepare_model():
     #pipe.vae.do_resize=False
     #pipe.vae.do_rescale=False
     #pipe.vae.do_convert_rgb=True
-    pipe.vae.vae_scale_factor=8
+    #pipe.vae.vae_scale_factor=8
 
     pipe.scheduler = sched
-    #pipe.scheduler =EDMDPMSolverMultistepScheduler.from_pretrained('John6666/uber-realistic-porn-merge-xl-urpmxl-v6final-sdxl', subfolder='scheduler', algorithm_type='sde-dpmsolver++')
     #pipe.vae=vae.to(torch.bfloat16)
     #pipe.unet=pipeX.unet
     #pipe.scheduler=EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", beta_start=0.00085, beta_end=0.012, steps_offset=1)
     #pipe.scheduler=EulerAncestralDiscreteScheduler.from_pretrained('ford442/RealVisXL_V5.0_BF16', subfolder='scheduler',beta_schedule="scaled_linear")
 
     pipe.to(device=device, dtype=torch.bfloat16)
-    #pipe.to(torch.bfloat16)
-    #apply_hidiffusion(pipe)
 
     #pipe.unet.set_default_attn_processor()
     pipe.vae.set_default_attn_processor()
@@ -196,10 +187,7 @@ def load_and_prepare_model():
     #print(f'UNET: {pipe.unet}')
     pipe.watermark=None
     pipe.safety_checker=None  
-    #pipe.to(torch.device("cuda:0"))
-    #pipe.vae.to(torch.bfloat16)
-    #pipe.to(device, torch.bfloat16)
-    #del pipeX
+
     
     # pipe.load_lora_weights("ford442/sdxl-vae-bf16", weight_name="LoRA/Fantasy_World_XL.safetensors", adapter_name="fantasy")
     pipe.load_lora_weights("ford442/sdxl-vae-bf16", weight_name="LoRA/skin_texture_style_v4.safetensors", adapter_name="skin")
@@ -207,13 +195,9 @@ def load_and_prepare_model():
     #pipe.set_adapters(["skin", "photo", "fantasy"], adapter_weights=[0.75, 0.25, 0.5])
     pipe.set_adapters(["skin"], adapter_weights=[0.5])
 
-    #sched = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config, beta_schedule="scaled_linear", algorithm_type="dpmsolver++")
     #sched = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config, beta_schedule="linear", algorithm_type="dpmsolver++")
-    #sched = DDIMScheduler.from_config(pipe.scheduler.config)
     return pipe
     
-# Preload and compile both models
-#models = {key: load_and_prepare_model(value) for key, value in MODEL_OPTIONS.items()}
 pipe = load_and_prepare_model()
 MAX_SEED = np.iinfo(np.int32).max
 
@@ -237,9 +221,8 @@ def save_image(img):
     img.save(unique_name,optimize=False,compress_level=0)
     return unique_name
 
-def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
+def randomize_seed_fn(seed: int) -> int:
+    seed = random.randint(0, MAX_SEED)
     return seed
 
 def uploadNote(prompt,num_inference_steps,guidance_scale,timestamp,denoise):
@@ -273,9 +256,9 @@ def generate_30(
     height: int = 768,
     guidance_scale: float = 4,
     num_inference_steps: int = 125,
-    randomize_seed: bool = False,
     use_resolution_binning: bool = True, 
     denoise: float = 0.3,
+    vae_scale: int = 8,
     progress=gr.Progress(track_tqdm=True)  # Add progress as a keyword argument
 ):
     #torch.backends.cudnn.benchmark = False
@@ -283,7 +266,8 @@ def generate_30(
     #gc.collect()
     #global models
     #pipe = models[model_choice]
-    seed = int(randomize_seed_fn(seed, randomize_seed))
+    pipe.vae.vae_scale_factor=vae_scale
+    seed = int(randomize_seed_fn(seed))
     generator = torch.Generator(device='cuda').manual_seed(seed)
     #prompt, negative_prompt = apply_style(style_selection, prompt, negative_prompt)
     options = {
@@ -307,7 +291,7 @@ def generate_30(
     uploadNote(prompt,num_inference_steps,guidance_scale,timestamp,denoise)
     batch_options = options.copy()
     rv_image = pipe(**batch_options).images[0]
-    sd_image_path = f"rv50_A_{seed}.png"
+    sd_image_path = f"rv50_A_{timestamp}.png"
     rv_image.save(sd_image_path,optimize=False,compress_level=0)
     upload_to_ftp(sd_image_path)
     image_paths = save_image(rv_image)
@@ -335,17 +319,18 @@ def generate_60(
     height: int = 768,
     guidance_scale: float = 4,
     num_inference_steps: int = 250,
-    randomize_seed: bool = False,
     use_resolution_binning: bool = True, 
     denoise: float = 0.3,
+    vae_scale: int = 8,
     progress=gr.Progress(track_tqdm=True)  # Add progress as a keyword argument
 ):
-    #torch.backends.cudnn.benchmark = True
+    #torch.backends.cudnn.benchmark = False
     #torch.cuda.empty_cache()
     #gc.collect()
     #global models
     #pipe = models[model_choice]
-    seed = int(randomize_seed_fn(seed, randomize_seed))
+    pipe.vae.vae_scale_factor=vae_scale
+    seed = int(randomize_seed_fn(seed))
     generator = torch.Generator(device='cuda').manual_seed(seed)
     #prompt, negative_prompt = apply_style(style_selection, prompt, negative_prompt)
     options = {
@@ -369,7 +354,7 @@ def generate_60(
     uploadNote(prompt,num_inference_steps,guidance_scale,timestamp,denoise)
     batch_options = options.copy()
     rv_image = pipe(**batch_options).images[0]
-    sd_image_path = f"rv50_A_{seed}.png"
+    sd_image_path = f"rv50_A_{timestamp}.png"
     rv_image.save(sd_image_path,optimize=False,compress_level=0)
     upload_to_ftp(sd_image_path)
     image_paths = save_image(rv_image)
@@ -397,17 +382,18 @@ def generate_90(
     height: int = 768,
     guidance_scale: float = 4,
     num_inference_steps: int = 250,
-    randomize_seed: bool = False,
     use_resolution_binning: bool = True, 
     denoise: float = 0.3,
+    vae_scale: int = 8,
     progress=gr.Progress(track_tqdm=True)  # Add progress as a keyword argument
 ):
-    #torch.backends.cudnn.benchmark = True
+    #torch.backends.cudnn.benchmark = False
     #torch.cuda.empty_cache()
     #gc.collect()
     #global models
     #pipe = models[model_choice]
-    seed = int(randomize_seed_fn(seed, randomize_seed))
+    pipe.vae.vae_scale_factor=vae_scale
+    seed = int(randomize_seed_fn(seed))
     generator = torch.Generator(device='cuda').manual_seed(seed)
     #prompt, negative_prompt = apply_style(style_selection, prompt, negative_prompt)
     options = {
@@ -431,7 +417,7 @@ def generate_90(
     uploadNote(prompt,num_inference_steps,guidance_scale,timestamp,denoise)
     batch_options = options.copy()
     rv_image = pipe(**batch_options).images[0]
-    sd_image_path = f"rv50_A_{seed}.png"
+    sd_image_path = f"rv50_A_{timestamp}.png"
     rv_image.save(sd_image_path,optimize=False,compress_level=0)
     upload_to_ftp(sd_image_path)
     image_paths = save_image(rv_image)
@@ -516,13 +502,6 @@ with gr.Blocks(theme=gr.themes.Origin(),css=css) as demo:
                     value="('deformed', 'distorted', 'disfigured':1.3),'not photorealistic':1.5, 'poorly drawn', 'bad anatomy', 'wrong anatomy', 'extra limb', 'missing limb', 'floating limbs', 'poorly drawn hands', 'poorly drawn feet', 'poorly drawn face':1.3, 'out of frame', 'extra limbs', 'bad anatomy', 'bad art', 'beginner',  'distorted face','amateur'",
                     visible=True,
                 )
-        seed = gr.Slider(
-            label="Seed",
-            minimum=0,
-            maximum=MAX_SEED,
-            step=1,
-            value=0,
-        )
         denoise = gr.Slider(
             label="Denoising Strength",
             minimum=0.0,
@@ -530,7 +509,13 @@ with gr.Blocks(theme=gr.themes.Origin(),css=css) as demo:
             step=0.01,
             value=0.3,
         )
-        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
+        vae_scale = gr.Slider(
+            label="VAE Scale Multiplier",
+            minimum=1,
+            maximum=16,
+            step=1,
+            value=8,
+        )
         with gr.Row():
             width = gr.Slider(
                 label="Width",
@@ -587,15 +572,14 @@ with gr.Blocks(theme=gr.themes.Origin(),css=css) as demo:
             negative_prompt,
             use_negative_prompt,
             style_selection,
-            seed,
             width,
             height,
             guidance_scale,
             num_inference_steps,
-            randomize_seed, 
-            denoise
+            denoise,
+            vae_scale,
         ],
-        outputs=[result, seed],
+        outputs=[result],
     )
     
     gr.on(
@@ -610,15 +594,14 @@ with gr.Blocks(theme=gr.themes.Origin(),css=css) as demo:
             negative_prompt,
             use_negative_prompt,
             style_selection,
-            seed,
             width,
             height,
             guidance_scale,
             num_inference_steps,
-            randomize_seed, 
-            denoise
+            denoise,
+            vae_scale,
         ],
-        outputs=[result, seed],
+        outputs=[result],
     )
     
     gr.on(
@@ -633,15 +616,14 @@ with gr.Blocks(theme=gr.themes.Origin(),css=css) as demo:
             negative_prompt,
             use_negative_prompt,
             style_selection,
-            seed,
             width,
             height,
             guidance_scale,
             num_inference_steps,
-            randomize_seed, 
-            denoise
+            denoise,
+            vae_scale,
         ],
-        outputs=[result, seed],
+        outputs=[result],
     )
 
     gr.Markdown("### REALVISXL V5.0")