feat(api): make LPW an image parameter

api/onnx_web/chain/blend_img2img.py

@@ -31,7 +31,10 @@ def blend_img2img(
         params.model,
         params.scheduler,
         job.get_device(),
+        params.lpw,
     )
+    if params.lpw:
+        pipe = pipe.img2img
 
     rng = torch.manual_seed(params.seed)
 

api/onnx_web/chain/blend_inpaint.py

@@ -64,7 +64,10 @@ def blend_inpaint(
             params.model,
             params.scheduler,
             job.get_device(),
+            params.lpw,
         )
+        if params.lpw:
+            pipe = pipe.inpaint
 
         latents = get_latents_from_seed(params.seed, size)
         rng = torch.manual_seed(params.seed)

api/onnx_web/chain/source_txt2img.py

@@ -32,8 +32,10 @@ def source_txt2img(
         )
 
     pipe = load_pipeline(
-        OnnxStableDiffusionPipeline, params.model, params.scheduler, job.get_device()
+        OnnxStableDiffusionPipeline, params.model, params.scheduler, job.get_device(), params.lpw
     )
+    if params.lpw:
+        pipe = pipe.text2img
 
     latents = get_latents_from_seed(params.seed, size)
     rng = torch.manual_seed(params.seed)

api/onnx_web/chain/upscale_outpaint.py

@@ -71,11 +71,13 @@ def upscale_outpaint(
             params.scheduler,
             job.get_device(),
         )
+        if params.lpw:
+            pipe = pipe.inpaint
 
         latents = get_tile_latents(full_latents, dims)
         rng = torch.manual_seed(params.seed)
 
-        result = pipe.inpaint(
+        result = pipe(
             image,
             mask,
             prompt,
@@ -96,7 +98,7 @@ def upscale_outpaint(
     margin_y = float(max(border.top, border.bottom))
     overlap = min(margin_x / source_image.width, margin_y / source_image.height)
 
-    if overlap > 0 and border.left == border.right and border.top == border.bottom:
+    if border.left == border.right and border.top == border.bottom:
         logger.debug("outpainting with an even border, using spiral tiling")
         output = process_tile_spiral(source_image, SizeChart.auto, 1, [outpaint], overlap=overlap)
     else:

api/onnx_web/diffusion/load.py

@@ -47,13 +47,13 @@ def get_tile_latents(
 
 
 def load_pipeline(
-    pipeline: DiffusionPipeline, model: str, scheduler: Any, device: DeviceParams
+    pipeline: DiffusionPipeline, model: str, scheduler: Any, device: DeviceParams, lpw: bool
 ):
     global last_pipeline_instance
     global last_pipeline_scheduler
     global last_pipeline_options
 
-    options = (pipeline, model, device.provider)
+    options = (pipeline, model, device.device, device.provider, lpw)
     if last_pipeline_instance is not None and last_pipeline_options == options:
         logger.debug("reusing existing diffusion pipeline")
         pipe = last_pipeline_instance
@@ -63,6 +63,11 @@ def load_pipeline(
         last_pipeline_scheduler = None
         run_gc()
 
+        if lpw:
+            custom_pipeline = "./onnx_web/diffusion/lpw_stable_diffusion_onnx.py"
+        else:
+            custom_pipeline = None
+
         logger.debug("loading new diffusion pipeline from %s", model)
         scheduler = scheduler.from_pretrained(
             model,
@@ -72,7 +77,7 @@ def load_pipeline(
         )
         pipe = pipeline.from_pretrained(
             model,
-            custom_pipeline="./onnx_web/diffusion/lpw_stable_diffusion_onnx.py",
+            custom_pipeline=custom_pipeline,
             provider=device.provider,
             provider_options=device.options,
             revision="onnx",

api/onnx_web/diffusion/run.py

@@ -25,14 +25,16 @@ def run_txt2img_pipeline(
     upscale: UpscaleParams,
 ) -> None:
     pipe = load_pipeline(
-        OnnxStableDiffusionPipeline, params.model, params.scheduler, job.get_device()
+        OnnxStableDiffusionPipeline, params.model, params.scheduler, job.get_device(), params.lpw
     )
+    if params.lpw:
+        pipe = pipe.text2img
 
     latents = get_latents_from_seed(params.seed, size)
     rng = torch.manual_seed(params.seed)
 
     progress = job.get_progress_callback()
-    result = pipe.text2img(
+    result = pipe(
         params.prompt,
         height=size.height,
         width=size.width,
@@ -72,12 +74,15 @@ def run_img2img_pipeline(
         params.model,
         params.scheduler,
         job.get_device(),
+        params.lpw
     )
+    if params.lpw:
+        pipe = pipe.img2img
 
     rng = torch.manual_seed(params.seed)
 
     progress = job.get_progress_callback()
-    result = pipe.img2img(
+    result = pipe(
         source_image,
         params.prompt,
         generator=rng,

api/onnx_web/params.py

@@ -86,10 +86,11 @@ class ImageParams:
         model: str,
         scheduler: Any,
         prompt: str,
-        negative_prompt: Optional[str],
         cfg: float,
         steps: int,
         seed: int,
+        negative_prompt: Optional[str] = None,
+        lpw: Optional[bool] = False,
     ) -> None:
         self.model = model
         self.scheduler = scheduler
@@ -98,6 +99,7 @@ class ImageParams:
         self.cfg = cfg
         self.seed = seed
         self.steps = steps
+        self.lpw = lpw or False
 
     def tojson(self) -> Dict[str, Optional[Param]]:
         return {
@@ -108,6 +110,7 @@ class ImageParams:
             "cfg": self.cfg,
             "seed": self.seed,
             "steps": self.steps,
+            "lpw": self.lpw,
         }
 
 

api/onnx_web/serve.py

@@ -171,6 +171,7 @@ def pipeline_from_request() -> Tuple[DeviceParams, ImageParams, Size]:
         device = available_platforms[0]
 
     # pipeline stuff
+    lpw = get_not_empty(request.args, "lpw", "false") == "true"
     model = get_not_empty(request.args, "model", get_config_value("model"))
     model_path = get_model_path(model)
     scheduler = get_from_map(
@@ -233,7 +234,7 @@ def pipeline_from_request() -> Tuple[DeviceParams, ImageParams, Size]:
     )
 
     params = ImageParams(
-        model_path, scheduler, prompt, negative_prompt, cfg, steps, seed
+        model_path, scheduler, prompt, negative_prompt, cfg, steps, seed, lpw=lpw
     )
     size = Size(width, height)
     return (device, params, size)