denoise_image = upscale_image(denoise_image, upscale, unit_resolution=32, min_size=min_size)

app.py

@@ -16,6 +16,8 @@ from PIL import Image
 from SUPIR.util import HWC3, upscale_image, fix_resize, convert_dtype, create_SUPIR_model, load_QF_ckpt
 from huggingface_hub import hf_hub_download
 
+max_64_bit_int = 2**63 - 1
+
 hf_hub_download(repo_id="laion/CLIP-ViT-bigG-14-laion2B-39B-b160k", filename="open_clip_pytorch_model.bin", local_dir="laion_CLIP-ViT-bigG-14-laion2B-39B-b160k")
 hf_hub_download(repo_id="camenduru/SUPIR", filename="sd_xl_base_1.0_0.9vae.safetensors", local_dir="yushan777_SUPIR")
 hf_hub_download(repo_id="camenduru/SUPIR", filename="SUPIR-v0F.ckpt", local_dir="yushan777_SUPIR")
@@ -57,7 +59,7 @@ def check_upload(input_image):
 
 def update_seed(is_randomize_seed, seed):
     if is_randomize_seed:
-        return random.randint(0, 2147483647)
+        return random.randint(0, max_64_bit_int)
     return seed
 
 def check(input_image):
@@ -206,6 +208,7 @@ def restore_in_Xmin(
         return [noisy_image, denoise_image], [denoise_image], None, None
 
     denoise_image = HWC3(np.array(denoise_image))
+    denoise_image = upscale_image(denoise_image, upscale, unit_resolution=32, min_size=min_size)
 
     # Allocation
     if allocation == 1:
@@ -331,8 +334,6 @@ def restore(
         elif model_select == 'v0-F':
             model.load_state_dict(ckpt_F, strict=False)
         model.current_model = model_select
-    input_image = upscale_image(input_image, upscale, unit_resolution=32,
-                                min_size=min_size)
 
     LQ = np.array(input_image) / 255.0
     LQ = np.power(LQ, gamma_correction)
@@ -474,7 +475,7 @@ with gr.Blocks() as interface:
         prompt = gr.Textbox(label="Image description", info="Help the AI understand what the image represents; describe as much as possible, especially the details we can't see on the original image; you can write in any language", value="", placeholder="A 33 years old man, walking, in the street, Santiago, morning, Summer, photorealistic", lines=3)
         prompt_hint = gr.HTML("You can use a <a href='"'https://huggingface.co/spaces/MaziyarPanahi/llava-llama-3-8b'"'>LlaVa space</a> to auto-generate the description of your image.")
         upscale = gr.Radio([["x1", 1], ["x2", 2], ["x3", 3], ["x4", 4], ["x5", 5], ["x6", 6], ["x7", 7], ["x8", 8], ["x9", 9], ["x10", 10]], label="Upscale factor", info="Resolution x1 to x10", value=2, interactive=True)
-        allocation = gr.Radio([["1 min", 1], ["2 min", 2], ["3 min", 3], ["4 min", 4], ["5 min", 5], ["6 min", 6], ["7 min", 7], ["8 min", 8], ["9 min (discouraged)", 9], ["10 min (discouraged)", 10]], label="GPU allocation time", info="lower=May abort run, higher=Quota penalty for next runs", value=6, interactive=True)
+        allocation = gr.Radio([["1 min", 1], ["2 min", 2], ["3 min", 3], ["4 min", 4], ["5 min", 5], ["6 min", 6], ["7 min", 7], ["8 min (discouraged)", 8], ["9 min (discouraged)", 9], ["10 min (discouraged)", 10]], label="GPU allocation time", info="lower=May abort run, higher=Quota penalty for next runs", value=6, interactive=True)
         output_format = gr.Radio([["As input", "input"], ["*.png", "png"], ["*.webp", "webp"], ["*.jpeg", "jpeg"], ["*.gif", "gif"], ["*.bmp", "bmp"]], label="Image format for result", info="File extention", value="input", interactive=True)
 
     with gr.Accordion("Pre-denoising (optional)", open=False):
@@ -526,13 +527,13 @@ with gr.Blocks() as interface:
                 spt_linear_s_stage2 = gr.Slider(label="Guidance Start", minimum=0.,
                                                 maximum=1., value=0., step=0.05)
             with gr.Column():
-                diff_dtype = gr.Radio([["fp32 (precision)", "fp32"], ["fp16 (medium)", "fp16"], ["bf16 (speed)", "bf16"]], label="Diffusion Data Type", value="fp16",
+                diff_dtype = gr.Radio([["fp32 (precision)", "fp32"], ["fp16 (medium)", "fp16"], ["bf16 (speed)", "bf16"]], label="Diffusion Data Type", value="fp32",
                                       interactive=True)
             with gr.Column():
                 ae_dtype = gr.Radio([["fp32 (precision)", "fp32"], ["bf16 (speed)", "bf16"]], label="Auto-Encoder Data Type", value="fp32",
                                     interactive=True)
         randomize_seed = gr.Checkbox(label = "\U0001F3B2 Randomize seed", value = True, info = "If checked, result is always different")
-        seed = gr.Slider(label="Seed", minimum=0, maximum=2**63 - 1, step=1, randomize=True)
+        seed = gr.Slider(label="Seed", minimum=0, maximum=max_64_bit_int, step=1, randomize=True)
         with gr.Group():
             param_setting = gr.Radio(["Quality", "Fidelity"], interactive=True, label="Presetting", value="Quality")
             restart_button = gr.Button(value="Apply presetting")