Update app.py

app.py

@@ -117,10 +117,13 @@ function custom(){
 # torch pipe
 
 pipe = AnimateDiffPipeline.from_pretrained(base, motion_adapter=adapter).to(device)
-pipe.scheduler = EulerAncestralDiscreteScheduler.from_pretrained(
-    base,
-    subfolder="scheduler",
+pipe.scheduler = DDIMScheduler(
+    clip_sample=False,
+    beta_start=0.00085,
+    beta_end=0.012,
     beta_schedule="linear",
+    timestep_spacing="trailing",
+    steps_offset=1
 )
 #pipe.unet.load_state_dict(load_file(hf_hub_download(repo, ckpt), device=device), strict=False)
 pipe.load_ip_adapter("h94/IP-Adapter", subfolder="models", weight_name="ip-adapter-plus_sd15.bin")
@@ -233,7 +236,7 @@ class Command:
         self.func = func
         self.data = data
 
-def parallel(*pairs):
+def parallel(timeout,*pairs):
     if len(pairs) == 0:
         return
     if len(pairs) == 1:
@@ -248,7 +251,7 @@ def parallel(*pairs):
         for pair in pairs:
             f = pair.pop(0)
             queue.put(Command(f, pair))
-        response = queue.get()
+        response = queue.get(timeout=timeout)
         return response
     
 # functionality
@@ -322,7 +325,7 @@ def calc(img,p1,p2,motion):
         negative_prompt=p2,
         height=height,
         width=width,
-        ip_adapter_image=img,
+        ip_adapter_image=img.convert("RGB"),
         num_inference_steps=step,
         guidance_scale=accu,
         num_frames=(fps*time)
@@ -351,7 +354,7 @@ def handle(*inp):
     ln = len(result)
 
     parargs = [[calc,*inp] for i in range(ln)]
-    out_pipe = parallel(parargs)
+    out_pipe = parallel(40,parargs)
     for i in range(ln):
         name = generate_random_string(12)+".png"
         export_to_gif(out_pipe[i].frames[0],name,fps=fps)