[pull] master from Comfy-Org:master

comfy/ldm/cascade/stage_a.py

@@ -136,16 +136,7 @@ def __init__(self, c, c_hidden):
             ops.Linear(c_hidden, c),
         )
 
-        self.gammas = nn.Parameter(torch.zeros(6), requires_grad=True)
-
-        # Init weights
-        def _basic_init(module):
-            if isinstance(module, nn.Linear) or isinstance(module, nn.Conv2d):
-                torch.nn.init.xavier_uniform_(module.weight)
-                if module.bias is not None:
-                    nn.init.constant_(module.bias, 0)
-
-        self.apply(_basic_init)
+        self.gammas = nn.Parameter(torch.zeros(6), requires_grad=False)
 
     def _norm(self, x, norm):
         return norm(x.permute(0, 2, 3, 1)).permute(0, 3, 1, 2)

comfy/model_management.py

@@ -1003,7 +1003,7 @@ def text_encoder_offload_device():
 def text_encoder_device():
     if args.gpu_only:
         return get_torch_device()
-    elif vram_state in (VRAMState.HIGH_VRAM, VRAMState.NORMAL_VRAM) or comfy.memory_management.aimdo_enabled:
+    elif vram_state in (VRAMState.HIGH_VRAM, VRAMState.NORMAL_VRAM, VRAMState.SHARED) or comfy.memory_management.aimdo_enabled:
         if should_use_fp16(prioritize_performance=False):
             return get_torch_device()
         else:

comfy/sd.py

@@ -455,7 +455,7 @@ def __init__(self, sd=None, device=None, config=None, dtype=None, metadata=None)
         self.output_channels = 3
         self.pad_channel_value = None
         self.process_input = lambda image: image * 2.0 - 1.0
-        self.process_output = lambda image: torch.clamp((image + 1.0) / 2.0, min=0.0, max=1.0)
+        self.process_output = lambda image: image.add_(1.0).div_(2.0).clamp_(0.0, 1.0)
         self.working_dtypes = [torch.bfloat16, torch.float32]
         self.disable_offload = False
         self.not_video = False
@@ -952,8 +952,8 @@ def decode(self, samples_in, vae_options={}):
             batch_number = max(1, batch_number)
 
             for x in range(0, samples_in.shape[0], batch_number):
-                samples = samples_in[x:x+batch_number].to(self.vae_dtype).to(self.device)
-                out = self.process_output(self.first_stage_model.decode(samples, **vae_options).to(self.output_device).to(dtype=self.vae_output_dtype()))
+                samples = samples_in[x:x + batch_number].to(device=self.device, dtype=self.vae_dtype)
+                out = self.process_output(self.first_stage_model.decode(samples, **vae_options).to(device=self.output_device, dtype=self.vae_output_dtype(), copy=True))
                 if pixel_samples is None:
                     pixel_samples = torch.empty((samples_in.shape[0],) + tuple(out.shape[1:]), device=self.output_device, dtype=self.vae_output_dtype())
                 pixel_samples[x:x+batch_number] = out