update

app.py

@@ -81,12 +81,7 @@ class CompVisDenoiser(K.external.CompVisDenoiser):
         return self.inner_model.apply_model(*args, **kwargs)
     
     def forward(self, input_0, input_1, sigma, **kwargs):
-        print("input_0.device:", input_0.device)
-        print("input_1.device:", input_1.device)
         c_out, c_in = [append_dims(x, input_0.ndim) for x in self.get_scalings(sigma)]
-        print("c_in.device:", c_in.device)
-        print("c_out.device:", c_out.device)
-        print("sigma.device:", sigma.device)
         # eps_0, eps_1 = self.get_eps(input_0 * c_in, input_1 * c_in, self.sigma_to_t(sigma), **kwargs)
         eps_0, eps_1 = self.get_eps(input_0 * c_in, self.sigma_to_t(sigma.cpu().float()).cuda(), **kwargs)
         
@@ -164,7 +159,7 @@ model_wrap = CompVisDenoiser(model)
 model_wrap_cfg = CFGDenoiser(model_wrap)
 null_token = model.get_learned_conditioning([""])
 
-@spaces.GPU(duration=20)
+@spaces.GPU(duration=200)
 def generate(
     input_image: Image.Image,
     instruction: str,
@@ -205,12 +200,8 @@ def generate(
         uncond["c_crossattn"] = [null_token.to(model.device)]
         uncond["c_concat"] = [torch.zeros_like(cond["c_concat"][0])]
         
-        print("cond['c_crossattn'][0].device:", cond["c_crossattn"][0].device)
-        print("cond['c_concat'][0].device:", cond["c_concat"][0].device)
-        print("uncond['c_crossattn'][0].device:", uncond["c_crossattn"][0].device)
-        print("uncond['c_concat'][0].device:", uncond["c_concat"][0].device)
 
-        sigmas = model_wrap.get_sigmas(steps).to(model.device)
+        sigmas = model_wrap.get_sigmas(steps)
 
         extra_args = {
             "cond": cond,
@@ -221,8 +212,6 @@ def generate(
         torch.manual_seed(seed)
         z_0 = torch.randn_like(cond["c_concat"][0]).to(model.device) * sigmas[0]
         z_1 = torch.randn_like(cond["c_concat"][0]).to(model.device) * sigmas[0]
-        print("z_0.device:", z_0.device)
-        print("z_1.device:", z_1.device)
         
         z_0, z_1, image_list, mask_list = sample_euler_ancestral(model_wrap_cfg, z_0, z_1, sigmas, height, width, extra_args=extra_args)