first modification

__init__.py

@@ -17,10 +17,13 @@ def common_ksampler(model, seed, steps, cfg, sampler_name, scheduler, positive,
     noise_mask = None
     device = model_management.get_torch_device()
 
-    if disable_noise:
-        noise = torch.zeros(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, device="cpu")
+    if "noise_sequence" in latent:
+        noise = latent["noise_sequence"]
     else:
-        noise = torch.randn(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, generator=torch.manual_seed(seed), device="cpu")
+        if disable_noise:
+            noise = torch.zeros(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, device="cpu")
+        else:
+            noise = torch.randn(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, generator=torch.manual_seed(seed), device="cpu")
 
     if "noise_mask_sequence" in latent:
         noise_mask_list = []
@@ -280,7 +283,9 @@ class DdimInversionSequence:
         model_management.load_model_gpu(model)
         context = context.to(device)
         samples = samples.to(device)
-        s = ddim_inversion(model, ddim_scheduler, samples, steps, context)
+        s = ddim_inversion(model, ddim_scheduler, samples, steps, context)[-1]
+        s = rearrange(s.squeeze(0), "c f h w -> f c h w")
+        s = s.cpu()
         return (s,)
 
 

sd.py

@@ -103,7 +103,8 @@ def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, e
     model = instantiate_from_config(model_config)
     model = load_model_weights(model, sd, verbose=False, load_state_dict_to=load_state_dict_to)
     model.model.diffusion_model = convert_unet_checkpoint(sd, OmegaConf.create({"model": model_config}))
+
     if fp16:
-        model = model.half()
+       model = model.half()
 
     return (ModelPatcher(model), clip, vae)

tuneavideo/models/unet.py

@@ -297,6 +297,9 @@ class UNet3DConditionModel(ModelMixin, ConfigMixin):
 
         sample = rearrange(x.unsqueeze(0), "b f c h w -> b c f h w")
 
+        sample = sample.type(self.dtype)
+        context = context.type(self.dtype)
+
         down_block_additional_residuals = None
         mid_block_additional_residual = None
 

tuneavideo/util.py

@@ -22,6 +22,7 @@ def next_step(model_output: Union[torch.FloatTensor, np.ndarray], timestep: int,
 def get_noise_pred_single(latents, t, context, unet):
     latents = rearrange(latents.squeeze(0), "c f h w -> f c h w")
     noise_pred = unet(latents, t.view(1), context=context)
+    noise_pred = rearrange(noise_pred.unsqueeze(0), "b f c h w -> b c f h w")
     return noise_pred