build: update requirements

imaginairy/cli/clickshell_mod.py

@@ -51,16 +51,16 @@ def invoke_(self, arg):  # pylint: disable=unused-argument
         return False
 
     invoke_ = update_wrapper(invoke_, command.callback)
-    invoke_.__name__ = "do_%s" % command.name
+    invoke_.__name__ = f"do_{command.name}"
     return invoke_
 
 
 class ModClickShell(ClickShell):
     def add_command(self, cmd, name):
         # Use the MethodType to add these as bound methods to our current instance
-        setattr(self, "do_%s" % name, get_method_type(mod_get_invoke(cmd), self))
-        setattr(self, "help_%s" % name, get_method_type(get_help(cmd), self))
-        setattr(self, "complete_%s" % name, get_method_type(get_complete(cmd), self))
+        setattr(self, f"do_{name}", get_method_type(mod_get_invoke(cmd), self))
+        setattr(self, f"help_{name}", get_method_type(get_help(cmd), self))
+        setattr(self, f"complete_{name}", get_method_type(get_complete(cmd), self))
 
 
 class ModShell(Shell):

imaginairy/enhancers/video_interpolation/rife/interpolate.py

@@ -310,12 +310,12 @@ def pad_image(img):
         if montage:
             write_buffer.put(np.concatenate((lastframe, lastframe), 1))
             for mid in output:
-                mid = (mid[0] * 255.0).byte().cpu().numpy().transpose(1, 2, 0)
+                mid = (mid[0] * 255.0).byte().cpu().numpy().transpose(1, 2, 0)  # type: ignore
                 write_buffer.put(np.concatenate((lastframe, mid[:h, :w]), 1))
         else:
             write_buffer.put(lastframe)
             for mid in output:
-                mid = (mid[0] * 255.0).byte().cpu().numpy().transpose(1, 2, 0)
+                mid = (mid[0] * 255.0).byte().cpu().numpy().transpose(1, 2, 0)  # type: ignore
                 write_buffer.put(mid[:h, :w])
         pbar.update(1)
         lastframe = frame

imaginairy/img_processors/control_modes.py

@@ -15,7 +15,7 @@ def create_canny_edges(img: "Tensor") -> "Tensor":
 
     img = torch.clamp((img + 1.0) / 2.0, min=0.0, max=1.0)
     img = einops.rearrange(img[0], "c h w -> h w c")
-    img = (255.0 * img).cpu().numpy().astype(np.uint8).squeeze()
+    img = (255.0 * img).cpu().numpy().astype(np.uint8).squeeze()  # type: ignore
     blurred = cv2.GaussianBlur(img, (5, 5), 0).astype(np.uint8)  # type: ignore
 
     if len(blurred.shape) > 2:

imaginairy/modules/diffusion/model.py

@@ -363,7 +363,7 @@ def make_attn(in_channels, attn_type="vanilla", attn_kwargs=None):
     if attn_type == "vanilla-xformers":
         # print(f"building MemoryEfficientAttnBlock with {in_channels} in_channels...")
         return MemoryEfficientAttnBlock(in_channels)
-    if type == "memory-efficient-cross-attn":
+    if attn_type == "memory-efficient-cross-attn":
         attn_kwargs["query_dim"] = in_channels
         return MemoryEfficientCrossAttentionWrapper(**attn_kwargs)
     if attn_type == "none":

imaginairy/modules/sgm/autoencoding/lpips/model/model.py

@@ -32,7 +32,7 @@ def __init__(self, input_nc=3, ndf=64, n_layers=3, use_actnorm=False):
         super().__init__()
         norm_layer = nn.BatchNorm2d if not use_actnorm else ActNorm
         if (
-            type(norm_layer) == functools.partial
+            type(norm_layer) == functools.partial  # noqa
         ):  # no need to use bias as BatchNorm2d has affine parameters
             use_bias = norm_layer.func != nn.BatchNorm2d
         else:

imaginairy/modules/sgm/diffusionmodules/model.py

@@ -308,7 +308,7 @@ def make_attn(in_channels, attn_type="vanilla", attn_kwargs=None):
         #     f"building MemoryEfficientAttnBlock with {in_channels} in_channels..."
         # )
         return MemoryEfficientAttnBlock(in_channels)
-    elif type == "memory-efficient-cross-attn":
+    elif attn_type == "memory-efficient-cross-attn":
         attn_kwargs["query_dim"] = in_channels
         return MemoryEfficientCrossAttentionWrapper(**attn_kwargs)
     elif attn_type == "none":

imaginairy/utils/animations.py

@@ -67,7 +67,7 @@ def make_bounce_animation(
 def _ensure_pillow_images(
     imgs: "List[Image.Image | LazyLoadingImage | torch.Tensor]",
 ) -> "List[Image.Image]":
-    converted_frames: "List[Image.Image]" = []
+    converted_frames: List[Image.Image] = []
     for frame in imgs:
         if isinstance(frame, torch.Tensor):
             converted_frames.append(model_latents_to_pillow_imgs(frame)[0])

imaginairy/weight_management/conversion.py

@@ -72,7 +72,7 @@ def could_convert(self, source_weights):
         return source_keys.issubset(self.all_valid_prefixes)
 
     def cast_weights(self, source_weights) -> dict[str, "Tensor"]:
-        converted_state_dict: dict[str, "Tensor"] = {}
+        converted_state_dict: dict[str, Tensor] = {}
         for source_key in source_weights:
             try:
                 source_prefix, suffix = source_key.rsplit(sep=".", maxsplit=1)