Merge pull request #3939 from vladmandic/dev

merge dev

CHANGELOG.md

@@ -3,7 +3,7 @@
 ## Update for 2025-05-15
 
 *Curious how your system is performing?*  
-Run a built-in benchmark and compare to over 15k unique results world-wide: (Benchmark data)[https://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html]!  
+Run a built-in benchmark and compare to over 15k unique results world-wide: [Benchmark data](https://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html)!  
 From slowest 0.02 it/s running on 6th gen CPU without acceleration up to 275+ it/s running on tuned GH100 system!  
 
 Also, since quantization is becoming a necessity for almost all new models, see comparison of different quantization methods available in SD.Next: [Quantization](https://vladmandic.github.io/sdnext-docs/Quantization/)  
@@ -26,6 +26,7 @@ Btw, last few releases have been smaller, but more regular so do check posts abo
   - Gradio: major cleanup and fixing defaults and ranges  
   - Pydantic: update to api types  
   - UI defaults: match correct prompt components  
+  - NNCF with ControlNet  
 
 ## Update for 2025-05-12
 

modules/control/run.py

@@ -433,7 +433,6 @@ def control_run(state: str = '',
         original_pipeline = shared.sd_model
         shared.sd_model = pipe
         sd_models.move_model(shared.sd_model, shared.device)
-        shared.sd_model.to(dtype=devices.dtype)
         debug_log(f'Control device={devices.device} dtype={devices.dtype}')
         sd_models.copy_diffuser_options(shared.sd_model, original_pipeline) # copy options from original pipeline
         sd_models.set_diffuser_options(shared.sd_model)

modules/face/instantid.py

@@ -61,7 +61,6 @@ def instant_id(p: processing.StableDiffusionProcessing, app, source_images, stre
     shared.sd_model.load_ip_adapter_instantid(face_adapter, scale=strength)
     shared.sd_model.set_ip_adapter_scale(strength)
     sd_models.move_model(shared.sd_model, devices.device) # move pipeline to device
-    shared.sd_model.to(dtype=devices.dtype)
 
     # pipeline specific args
     if p.all_prompts is None or len(p.all_prompts) == 0:

modules/intel/ipex/diffusers.py

@@ -61,12 +61,33 @@ def hidream_rope(pos: torch.Tensor, dim: int, theta: int) -> torch.Tensor:
     return out.to(return_device, dtype=torch.float32)
 
 
+def get_1d_sincos_pos_embed_from_grid(embed_dim, pos, output_type="np"):
+    if output_type == "np":
+        return diffusers.models.embeddings.get_1d_sincos_pos_embed_from_grid_np(embed_dim=embed_dim, pos=pos)
+    if embed_dim % 2 != 0:
+        raise ValueError("embed_dim must be divisible by 2")
+
+    omega = torch.arange(embed_dim // 2, device=pos.device, dtype=torch.float32)
+    omega /= embed_dim / 2.0
+    omega = 1.0 / 10000**omega  # (D/2,)
+
+    pos = pos.reshape(-1)  # (M,)
+    out = torch.outer(pos, omega)  # (M, D/2), outer product
+
+    emb_sin = torch.sin(out)  # (M, D/2)
+    emb_cos = torch.cos(out)  # (M, D/2)
+
+    emb = torch.concat([emb_sin, emb_cos], dim=1)  # (M, D)
+    return emb
+
+
 def ipex_diffusers(device_supports_fp64=False, can_allocate_plus_4gb=False):
     diffusers.utils.torch_utils.fourier_filter = fourier_filter
     if not device_supports_fp64:
         # get around lazy imports
         from diffusers.models import transformers as diffusers_transformers # pylint: disable=import-error, unused-import # noqa: F401
         from diffusers.models import controlnets as diffusers_controlnets # pylint: disable=import-error, unused-import # noqa: F401
+        diffusers.models.embeddings.get_1d_sincos_pos_embed_from_grid = get_1d_sincos_pos_embed_from_grid
         diffusers.models.embeddings.FluxPosEmbed = FluxPosEmbed
         diffusers.models.transformers.transformer_flux.FluxPosEmbed = FluxPosEmbed
         diffusers.models.controlnets.controlnet_flux.FluxPosEmbed = FluxPosEmbed

scripts/example.py

@@ -110,7 +110,6 @@ class Script(scripts.Script):
         sd_models.copy_diffuser_options(shared.sd_model, orig_pipeline) # copy options from original pipeline
         sd_models.set_diffuser_options(shared.sd_model) # set all model options such as fp16, offload, etc.
         sd_models.move_model(shared.sd_model, devices.device) # move pipeline to device
-        shared.sd_model.to(dtype=devices.dtype)
 
         # if pipeline also needs a specific type, you can set it here, but not commonly needed
         # shared.sd_model = sd_models.set_diffuser_pipe(shared.sd_model, sd_models.DiffusersTaskType.IMAGE_2_IMAGE)