Merge branch 'main' of https://github.com/Stability-AI/stability-sdk into feature/face-enhance-v1

Author: enzymezoo-code

.github/workflows/publish-to-pypi.yml

@@ -43,4 +43,4 @@ jobs:
       - name: Publish to PyPI
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
-          password: ${{ secrets.PYPI_API_TOKEN }}
+          password: ${{ secrets.PYPI_API_TOKEN }}

.github/workflows/unit_testing.yaml

@@ -1,6 +1,6 @@
 name: run unit tests
 
-on: [push, workflow_dispatch]
+on: [push, workflow_dispatch, pull_request]
 
 jobs:
   test:

.gitignore

@@ -7,4 +7,4 @@ pyenv/
 *venv/
 .env
 generation-*.pb.json
-Pipfile*
+Pipfile*

LICENSE

@@ -1,5 +1,5 @@
 
-    Copyright (c) Stability.ai, All rights reserved.
+    Copyright (c) Stability AI Ltd, All rights reserved.
 
     Permission is hereby granted, free of charge, to any person obtaining a copy
     of this software and associated documentation files (the "Software"), to deal

README.md

@@ -26,7 +26,7 @@ Install the [PyPI](https://pypi.org/project/stability-sdk/) package via:
 
 Then to invoke:
 
-`python3 -m stability_sdk generate -W 512 -H 512 "A stunning house."`
+`python3 -m stability_sdk generate -W 1024 -H 1024 "A stunning house."`
 
 It will generate and put PNGs in your current directory.
 
@@ -48,22 +48,23 @@ Be sure to check out [Platform](https://platform.stability.ai) for comprehensive
 ## Command line usage
 
 ```
-usage: python -m stability_sdk generate [-h] [--height HEIGHT] [--width WIDTH] [--start_schedule START_SCHEDULE]
-                 [--end_schedule END_SCHEDULE] [--cfg_scale CFG_SCALE] [--sampler SAMPLER]
-                 [--steps STEPS] [--seed SEED] [--prefix PREFIX] [--engine ENGINE]
-                 [--num_samples NUM_SAMPLES] [--artifact_types ARTIFACT_TYPES]
-                 [--no-store] [--show] [--init_image INIT_IMAGE] [--mask_image MASK_IMAGE]
-                 [prompt ...]
+usage: python -m stability_sdk generate [-h] [--height HEIGHT] [--width WIDTH] 
+                [--start_schedule START_SCHEDULE] [--end_schedule END_SCHEDULE] 
+                [--cfg_scale CFG_SCALE] [--sampler SAMPLER] [--steps STEPS] 
+                [--style_preset STYLE_PRESET] [--seed SEED] [--prefix PREFIX] [--engine ENGINE]
+                [--num_samples NUM_SAMPLES] [--artifact_types ARTIFACT_TYPES]
+                [--no-store] [--show] [--init_image INIT_IMAGE] [--mask_image MASK_IMAGE]
+                [prompt ...]
 
 positional arguments:
   prompt
 
 options:
   -h, --help            show this help message and exit
   --height HEIGHT, -H HEIGHT
-                        [512] height of image
+                        [1024] height of image
   --width WIDTH, -W WIDTH
-                        [512] width of image
+                        [1024] width of image
   --start_schedule START_SCHEDULE
                         [0.5] start schedule for init image (must be greater than 0; 1 is full strength
                         text prompt, no trace of image)
@@ -76,6 +77,10 @@ options:
                         k_dpm_2_ancestral, k_lms, k_dpmpp_2m, k_dpmpp_2s_ancestral)
   --steps STEPS, -s STEPS
                         [auto] number of steps
+  --style_preset STYLE_PRESET
+                        [none] (3d-model, analog-film, anime, cinematic, comic-book, digital-art, enhance, 
+                        fantasy-art, isometric, line-art, low-poly, modeling-compound, neon-punk, origami, 
+                        photographic, pixel-art, tile-texture)
   --seed SEED, -S SEED  random seed to use
   --prefix PREFIX, -p PREFIX
                         output prefixes for artifacts

pyproject.toml

@@ -4,4 +4,4 @@ build-backend = "setuptools.build_meta"
 
 [tool.pytest.ini_options]
 log_cli = true
-log_cli_level = "DEBUG"
+log_cli_level = "DEBUG"

setup.py

@@ -10,7 +10,7 @@
 
 setup(
     name='stability-sdk',
-    version='0.9.0',
+    version='0.8.4',
     author='Stability AI',
     author_email='support@stability.ai',
     maintainer='Stability AI',

src/stability_sdk/animation.py

@@ -62,7 +62,7 @@ class BasicSettings(param.Parameterized):
         check_on_set=False, # allow old and new models without raising ValueError
         objects=[
             "stable-diffusion-v1-5", "stable-diffusion-512-v2-1", "stable-diffusion-768-v2-1", 
-            "stable-diffusion-depth-v2-0", "stable-diffusion-xl-beta-v2-2-2",
+            "stable-diffusion-depth-v2-0", "stable-diffusion-xl-beta-v2-2-2", "stable-diffusion-xl-1024-v1-0",
             "custom"
         ]
     )
@@ -332,6 +332,10 @@ def __init__(
         # configure Api to retry on classifier obfuscations
         self.api._retry_obfuscation = True
 
+        # two stage 1024 model requires longer timeout
+        if self.args.model.startswith('stable-diffusion-xl-1024'):
+            self.api._request_timeout = 120.0
+
         # create output directory
         if self.out_dir is not None:
             os.makedirs(self.out_dir, exist_ok=True)

src/stability_sdk/api.py

@@ -69,7 +69,7 @@ def __init__(
             host: str="", 
             api_key: str=None, 
             stub: generation_grpc.GenerationServiceStub=None,
-            generate_engine_id: str="stable-diffusion-xl-beta-v2-2-2",
+            generate_engine_id: str="stable-diffusion-xl-1024-v1-0",
             inpaint_engine_id: str="stable-inpainting-512-v2-0",
             interpolate_engine_id: str="interpolation-server-v1",
             transform_engine_id: str="transform-server-v1",
@@ -104,8 +104,8 @@ def generate(
         self,
         prompts: List[str], 
         weights: List[float], 
-        width: int = 512, 
-        height: int = 512, 
+        width: int = 1024, 
+        height: int = 1024, 
         steps: Optional[int] = None,
         seed: Union[Sequence[int], int] = 0,
         samples: int = 1,

src/stability_sdk/client.py

@@ -95,7 +95,7 @@ def __init__(
         self,
         host: str = "grpc.stability.ai:443",
         key: str = "",
-        engine: str = "stable-diffusion-xl-beta-v2-2-2",
+        engine: str = "stable-diffusion-xl-1024-v1-0",
         upscale_engine: str = "esrgan-v1-x2plus",
         enhance_engine: str = "face-enhance-v1",
         verbose: bool = False,
@@ -157,8 +157,8 @@ def generate(
         prompt: Union[str, List[str], generation.Prompt, List[generation.Prompt]],
         init_image: Optional[Image.Image] = None,
         mask_image: Optional[Image.Image] = None,
-        height: int = 512,
-        width: int = 512,
+        height: int = 1024,
+        width: int = 1024,
         start_schedule: float = 1.0,
         end_schedule: float = 0.01,
         cfg_scale: float = 7.0,
@@ -175,6 +175,10 @@ def generate(
         guidance_models: List[str] = None,
         upscale: Union[bool, Dict[str, Any]] = False,
         enhance: Union[bool, Dict[str, Any]] = False,
+        adapter_type: generation.T2IAdapter = None,
+        adapter_strength: float = 0.4,
+        adapter_init_type: generation.T2IAdapterInit = generation.T2IADAPTERINIT_IMAGE,
+        style_preset: Optional[str] = None
     ) -> Generator[generation.Answer, None, None]:
         """
         Generate images from a prompt.
@@ -200,6 +204,11 @@ def generate(
         :param guidance_models: Models to use for guidance.
         :param upscale: Whether to upscale the generated images. Can also pass a dictionary of upscale arguments. See client.upscale for supported values.
         :param enhance: Whether to enhance the generated images. Can also pass a dictionary of enhance arguments. See client._make_enhance_request for supported values.
+        :param adapter_type: T2I adapter type, if any.
+        :param adapter_strength: Float between 0, 1 representing the proportion of unet passes into which we inject adapter weights
+        :param adapter_init_type: If T2IADAPTERINIT_IMAGE then init_image is converted into an initialising image corresponding to the adapter_type. i.e.
+        a sketch/depthmap/canny edge. If T2IADAPTERINIT_ADAPTER_IMAGE, then the init_image is treated as already a a sketch/depthmap/canny edge.
+        :param style_preset: Style preset name to use (see https://platform.stability.ai/docs/api-reference#tag/v1generation)
         :return: Generator of Answer objects.
         """
 
@@ -220,7 +229,11 @@ def generate(
                 guidance_cuts = guidance_cuts,
                 guidance_strength = guidance_strength,
                 guidance_prompt = guidance_prompt,
-                guidance_models = guidance_models
+                guidance_models = guidance_models,
+                adapter_type = adapter_type,
+                adapter_strength = adapter_strength,
+                adapter_init_type = adapter_init_type,
+                style_preset = style_preset
             )
 
         if not upscale and not enhance:
@@ -378,6 +391,10 @@ def _make_generate_request(
         guidance_strength: Optional[float] = None,
         guidance_prompt: Union[str, generation.Prompt] = None,
         guidance_models: List[str] = None,
+        adapter_type: generation.T2IAdapter = None,
+        adapter_strength: float = 0.4,
+        adapter_init_type: generation.T2IAdapterInit = generation.T2IADAPTERINIT_IMAGE,
+        style_preset: Optional[str] = None
     ):
         """
         Create a generate request
@@ -435,8 +452,7 @@ def _make_generate_request(
                 raise ValueError("guidance_prompt must be a string or Prompt object")
         if guidance_strength == 0.0:
             guidance_strength = None
-
-
+        
         # Build our CLIP parameters
         if guidance_preset is not generation.GUIDANCE_PRESET_NONE:
             # to do: make it so user can override this
@@ -464,6 +480,12 @@ def _make_generate_request(
                 ],
             )
 
+        adapter_parameters = generation.T2IAdapterParameter(
+            adapter_type = adapter_type,
+            adapter_strength = adapter_strength,
+            adapter_init_type = adapter_init_type,
+        )
+
         transform=None
         if sampler:
             transform=generation.TransformType(diffusion=sampler)
@@ -475,17 +497,25 @@ def _make_generate_request(
             seed=seed,
             steps=steps,
             samples=samples,
+            adapter=adapter_parameters,
             parameters=[generation.StepParameter(**step_parameters)],
         )
 
         request_id = str(uuid.uuid4())
         engine_id = self.engine
 
+        if style_preset and style_preset.lower() != 'none':
+            extras = Struct()
+            extras.update({ '$IPC': { "preset": style_preset } })
+        else:
+            extras = None
+
         rq = generation.Request(
             engine_id=engine_id,
             request_id=request_id,
             prompt=prompts,
             image=image_parameters,
+            extras=extras
         )
 
         return rq
@@ -641,9 +671,10 @@ def run_request(self,
             yield answer
             start = time.time()
 
-def process_cli(logger: logging.Logger = None,
-                warn_client_call_deprecated: bool = True,
-                ):
+def process_cli(
+    logger: logging.Logger = None,
+    warn_client_call_deprecated: bool = True,
+):
     if not logger:
         logger = logging.getLogger(__name__)
         logger.setLevel(level=logging.INFO)
@@ -742,10 +773,10 @@ def process_cli(logger: logging.Logger = None,
 
     parser_generate = subparsers.add_parser('generate')
     parser_generate.add_argument(
-        "--height", "-H", type=int, default=512, help="[512] height of image"
+        "--height", "-H", type=int, default=1024, help="[1024] height of image"
     )
     parser_generate.add_argument(
-        "--width", "-W", type=int, default=512, help="[512] width of image"
+        "--width", "-W", type=int, default=1024, help="[1024] width of image"
     )
     parser_generate.add_argument(
         "--start_schedule",
@@ -773,6 +804,7 @@ def process_cli(logger: logging.Logger = None,
     )
     parser_generate.add_argument(
         "--seed", "-S", type=int, default=0, help="random seed to use")
+    parser_generate.add_argument("--style_preset", type=str, help="style preset name")
     parser_generate.add_argument(
         "--prefix",
         "-p",
@@ -799,7 +831,7 @@ def process_cli(logger: logging.Logger = None,
         "-e",
         type=str,
         help="engine to use for inference",
-        default="stable-diffusion-xl-beta-v2-2-2",
+        default="stable-diffusion-xl-1024-v1-0",
     )
     parser_generate.add_argument(
         "--init_image",
@@ -908,6 +940,7 @@ def process_cli(logger: logging.Logger = None,
             "mask_image": args.mask_image,
             "upscale": upscale,
             "enhance": enhance,
+            "style_preset": args.style_preset,
         }
 
         if args.sampler: