fix(runner): improve 'num_inference_steps' logic (livepeer#205)

This commit prevents a Key Error from being thrown when the pipelines
are called directly.

runner/app/pipelines/image_to_image.py

@@ -197,7 +197,6 @@ def __call__(
         self, prompt: str, image: PIL.Image, **kwargs
     ) -> Tuple[List[PIL.Image], List[Optional[bool]]]:
         seed = kwargs.pop("seed", None)
-        num_inference_steps = kwargs.get("num_inference_steps", None)
         safety_check = kwargs.pop("safety_check", True)
 
         if seed is not None:
@@ -206,7 +205,9 @@ def __call__(
             elif isinstance(seed, list):
                 kwargs["generator"] = [torch.Generator(get_torch_device()).manual_seed(s) for s in seed]
 
-        if num_inference_steps is None or num_inference_steps < 1:
+        if "num_inference_steps" in kwargs and (
+            kwargs["num_inference_steps"] is None or kwargs["num_inference_steps"] < 1
+        ):
             del kwargs["num_inference_steps"]
 
         if self.model_id in ["stabilityai/sdxl-turbo", "stabilityai/sd-turbo"]:

runner/app/pipelines/image_to_video.py

@@ -111,7 +111,6 @@ def __call__(
         self, image: PIL.Image, **kwargs
     ) -> Tuple[List[PIL.Image], List[Optional[bool]]]:
         seed = kwargs.pop("seed", None)
-        num_inference_steps = kwargs.get("num_inference_steps", None)
         safety_check = kwargs.pop("safety_check", True)
 
         if "decode_chunk_size" not in kwargs:
@@ -128,7 +127,9 @@ def __call__(
                     torch.Generator(get_torch_device()).manual_seed(s) for s in seed
                 ]
 
-        if num_inference_steps is None or num_inference_steps < 1:
+        if "num_inference_steps" in kwargs and (
+            kwargs["num_inference_steps"] is None or kwargs["num_inference_steps"] < 1
+        ):
             del kwargs["num_inference_steps"]
 
         if safety_check:

runner/app/pipelines/text_to_image.py

@@ -226,7 +226,6 @@ def __call__(
         self, prompt: str, **kwargs
     ) -> Tuple[List[PIL.Image], List[Optional[bool]]]:
         seed = kwargs.pop("seed", None)
-        num_inference_steps = kwargs.get("num_inference_steps", None)
         safety_check = kwargs.pop("safety_check", True)
 
         if seed is not None:
@@ -239,7 +238,9 @@ def __call__(
                     torch.Generator(get_torch_device()).manual_seed(s) for s in seed
                 ]
 
-        if num_inference_steps is None or num_inference_steps < 1:
+        if "num_inference_steps" in kwargs and (
+            kwargs["num_inference_steps"] is None or kwargs["num_inference_steps"] < 1
+        ):
             del kwargs["num_inference_steps"]
 
         if (

runner/app/pipelines/upscale.py

@@ -118,7 +118,6 @@ def __call__(
         self, prompt: str, image: PIL.Image, **kwargs
     ) -> Tuple[List[PIL.Image], List[Optional[bool]]]:
         seed = kwargs.pop("seed", None)
-        num_inference_steps = kwargs.get("num_inference_steps", None)
         safety_check = kwargs.pop("safety_check", True)
 
         if seed is not None:
@@ -131,8 +130,10 @@ def __call__(
                     torch.Generator(get_torch_device()).manual_seed(s) for s in seed
                 ]
 
-        if num_inference_steps is None or num_inference_steps < 1:
-            kwargs.pop("num_inference_steps", None)
+        if "num_inference_steps" in kwargs and (
+            kwargs["num_inference_steps"] is None or kwargs["num_inference_steps"] < 1
+        ):
+            del kwargs["num_inference_steps"]
 
         # trying differnt configs of promp_embed for different models
         try: