update
Browse files- handler.py +9 -11
handler.py
CHANGED
|
@@ -2,16 +2,14 @@ from typing import Dict, Any
|
|
| 2 |
import torch
|
| 3 |
from diffusers import StableDiffusionInstructPix2PixPipeline, EulerAncestralDiscreteScheduler
|
| 4 |
from PIL import Image
|
| 5 |
-
import
|
| 6 |
-
from
|
| 7 |
-
|
| 8 |
-
model_id = "timbrooks/instruct-pix2pix"
|
| 9 |
|
| 10 |
class EndpointHandler:
|
| 11 |
|
| 12 |
-
def __init__(self):
|
| 13 |
model_id = "timbrooks/instruct-pix2pix"
|
| 14 |
-
self.model = StableDiffusionInstructPix2PixPipeline.from_pretrained(
|
| 15 |
self.model.scheduler = EulerAncestralDiscreteScheduler.from_config(self.model.scheduler.config)
|
| 16 |
|
| 17 |
def __call__(self, data: Dict[str, Any]) -> Image:
|
|
@@ -20,6 +18,7 @@ class EndpointHandler:
|
|
| 20 |
data (dict):
|
| 21 |
The payload with the input image, text prompt, color code, and optional parameters.
|
| 22 |
"""
|
|
|
|
| 23 |
input_image = self.load_process_image(data['image_path'])
|
| 24 |
text_prompt = data['text_prompt']
|
| 25 |
color_code = data['color_code']
|
|
@@ -36,10 +35,9 @@ class EndpointHandler:
|
|
| 36 |
num_inference_steps=50,
|
| 37 |
guidance_scale=guidance_scale,
|
| 38 |
image_guidance_scale=image_guidance_scale,
|
| 39 |
-
|
| 40 |
).images
|
| 41 |
|
| 42 |
-
return
|
| 43 |
|
| 44 |
def load_process_image(self, image_path):
|
| 45 |
image = Image.open(image_path)
|
|
@@ -53,8 +51,8 @@ class EndpointHandler:
|
|
| 53 |
result_prompt = f"{text_prompt}{coloring_prompt}"
|
| 54 |
return result_prompt
|
| 55 |
|
| 56 |
-
def hex_to_name(hex_color):
|
| 57 |
-
rgb_tuple = tuple(int(
|
| 58 |
names = []
|
| 59 |
rgb_values = []
|
| 60 |
|
|
@@ -64,4 +62,4 @@ class EndpointHandler:
|
|
| 64 |
|
| 65 |
kdt_db = KDTree(rgb_values)
|
| 66 |
distance, index = kdt_db.query(rgb_tuple)
|
| 67 |
-
return names[index]
|
|
|
|
| 2 |
import torch
|
| 3 |
from diffusers import StableDiffusionInstructPix2PixPipeline, EulerAncestralDiscreteScheduler
|
| 4 |
from PIL import Image
|
| 5 |
+
from webcolors import CSS3_HEX_TO_NAMES, hex_to_rgb
|
| 6 |
+
from scipy.spatial import KDTree
|
|
|
|
|
|
|
| 7 |
|
| 8 |
class EndpointHandler:
|
| 9 |
|
| 10 |
+
def __init__(self, path=""):
|
| 11 |
model_id = "timbrooks/instruct-pix2pix"
|
| 12 |
+
self.model = StableDiffusionInstructPix2PixPipeline.from_pretrained(path, torch_dtype=torch.float16, safety_checker=None).to("cuda")
|
| 13 |
self.model.scheduler = EulerAncestralDiscreteScheduler.from_config(self.model.scheduler.config)
|
| 14 |
|
| 15 |
def __call__(self, data: Dict[str, Any]) -> Image:
|
|
|
|
| 18 |
data (dict):
|
| 19 |
The payload with the input image, text prompt, color code, and optional parameters.
|
| 20 |
"""
|
| 21 |
+
|
| 22 |
input_image = self.load_process_image(data['image_path'])
|
| 23 |
text_prompt = data['text_prompt']
|
| 24 |
color_code = data['color_code']
|
|
|
|
| 35 |
num_inference_steps=50,
|
| 36 |
guidance_scale=guidance_scale,
|
| 37 |
image_guidance_scale=image_guidance_scale,
|
|
|
|
| 38 |
).images
|
| 39 |
|
| 40 |
+
return images
|
| 41 |
|
| 42 |
def load_process_image(self, image_path):
|
| 43 |
image = Image.open(image_path)
|
|
|
|
| 51 |
result_prompt = f"{text_prompt}{coloring_prompt}"
|
| 52 |
return result_prompt
|
| 53 |
|
| 54 |
+
def hex_to_name(self, hex_color):
|
| 55 |
+
rgb_tuple = tuple(int(hex_color[i:i+2], 16) for i in (1, 3, 5))
|
| 56 |
names = []
|
| 57 |
rgb_values = []
|
| 58 |
|
|
|
|
| 62 |
|
| 63 |
kdt_db = KDTree(rgb_values)
|
| 64 |
distance, index = kdt_db.query(rgb_tuple)
|
| 65 |
+
return names[index]
|