Merge pull request #152 from intel/update-branch

chores: update default model precision in text to image generation microservice (#406)

Author: gooishin

usecases/ai/microservices/text-to-image/flux-dev/backend/server.py

@@ -2,7 +2,6 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import time
-import openvino.runtime as ov_runtime
 import openvino_genai as ov_genai
 import openvino as ov
 import torch
@@ -105,7 +104,7 @@ def convert_models(self):
     @staticmethod
     def get_device(user_device=None):
         try:
-            ov_core = ov_runtime.Core()
+            ov_core = ov.Core()
             available_devices = [device.upper() for device in ov_core.available_devices]  # Normalize device names
             print(f"Available devices: {available_devices}")
 

usecases/ai/microservices/text-to-image/flux-schnell/backend/client.py

@@ -98,9 +98,9 @@ def main():
         "/pipeline/run",
         {
             "prompt": "A raccoon trapped inside a glass jar full of colorful candies, the background is steamy with vivid colors",
-            "width": 512,  # Additional parameter: width
-            "height": 512,  # Additional parameter: height
-            "num_inference_steps": 5  # Additional parameter: num_inference_steps
+            "width": 256,  # Additional parameter: width
+            "height": 256,  # Additional parameter: height
+            "num_inference_steps": 1  # Additional parameter: num_inference_steps
         },
     )
     if not response:

usecases/ai/microservices/text-to-image/flux-schnell/backend/server.py

@@ -2,7 +2,6 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import time
-import openvino.runtime as ov_runtime
 import openvino_genai as ov_genai
 import openvino as ov
 import torch
@@ -97,15 +96,15 @@ def convert_models(self):
         if not self.model_dir.exists():
             print(f"Downloading model: {self.model_name} to {self.model_dir}...")
             additional_args = {}
-            additional_args.update({"weight-format": "int4", "group-size": "64", "ratio": "1.0"})
+            additional_args.update({"weight-format": "int8", "group-size": "64", "ratio": "1.0"})
             optimum_cli(self.model_name, self.model_dir, additional_args=additional_args)
             # optimum_cli(self.model_name, self.model_dir)
         print("Model conversion completed.")
 
     @staticmethod
     def get_device(user_device=None):
         try:
-            ov_core = ov_runtime.Core()
+            ov_core = ov.Core()
             available_devices = [device.upper() for device in ov_core.available_devices]  # Normalize device names
             print(f"Available devices: {available_devices}")
 

usecases/ai/microservices/text-to-image/stable-diffusion-v3.5/backend/server.py

@@ -2,7 +2,6 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import time
-import openvino.runtime as ov_runtime
 import openvino_genai as ov_genai
 import openvino as ov
 import torch
@@ -102,7 +101,7 @@ def convert_models(self):
     @staticmethod
     def get_device(user_device=None):
         try:
-            ov_core = ov_runtime.Core()
+            ov_core = ov.Core()
             available_devices = [device.upper() for device in ov_core.available_devices]  # Normalize device names
             print(f"Available devices: {available_devices}")
 

usecases/ai/microservices/text-to-image/stable-diffusion-v3/backend/server.py

@@ -8,7 +8,7 @@
 from pydantic import BaseModel
 
 import torch
-import openvino.runtime as ov_runtime
+import openvino as ov
 
 from sd3_helper import (
     get_pipeline_options,
@@ -106,7 +106,7 @@ def convert_models(self):
     @staticmethod
     def get_device(user_device=None):
         try:
-            ov_core = ov_runtime.Core()
+            ov_core = ov.Core()
             available_devices = [device.upper() for device in ov_core.available_devices]  # Normalize device names
             print(f"Available devices: {available_devices}")
 

usecases/ai/microservices/text-to-image/stable-diffusion-xl/backend/server.py

@@ -2,7 +2,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import time
-import openvino.runtime as ov_runtime
+import openvino as ov
 import random
 import torch
 import gc
@@ -105,7 +105,7 @@ def convert_models(self):
     @staticmethod
     def get_device(user_device=None):
         try:
-            ov_core = ov_runtime.Core()
+            ov_core = ov.Core()
             available_devices = [device.upper() for device in ov_core.available_devices]  # Normalize device names
             print(f"Available devices: {available_devices}")