update demo to SDXL Turbo

app.py

@@ -150,13 +150,13 @@ with block:
                 "
               >
                 <h1 style="font-weight: 900; margin-bottom: 7px;">
-                  Fair Compute Demo
+                  SDXL Turbo by Fair Compute
                 </h1>
               </div>
               <p style="margin-bottom: 10px; font-size: 94%">
-                A demo of a popular Stable Diffusion neural network that runs on regular computers.
-                Run AI models on your computer and avoid paying egregious cloud fees
-                This demo is powered by the <a href="https://faircompute.com/">FairCompute</a> platform.
+                SDXL Turbo model for generating high-quality images.<br/>
+                Model: <a href="https://huggingface.co/stabilityai/sdxl-turbo">https://huggingface.co/stabilityai/sdxl-turbo</a><br/>
+                <br/>
               </p>
             </div>
         """
@@ -191,15 +191,9 @@ with block:
 
         gr.HTML(
             """
-                <div class="footer">
-                    <p>Model by <a href="https://huggingface.co/CompVis" style="text-decoration: underline;" target="_blank">CompVis</a> and <a href="https://runwayml.com/" style="text-decoration: underline;" target="_blank">Runway</a> powered by <a href="https://faircompute.com/" style="text-decoration: underline;" target="_blank">FairCompute</a> platform
-                    </p>
-                </div>
                 <div class="acknowledgments">
-                    <p><h4>LICENSE</h4>
-The model is licensed with a <a href="https://huggingface.co/spaces/CompVis/stable-diffusion-license" style="text-decoration: underline;" target="_blank">CreativeML Open RAIL-M</a> license. The authors claim no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in this license. The license forbids you from sharing any content that violates any laws, produce any harm to a person, disseminate any personal information that would be meant for harm, spread misinformation and target vulnerable groups. For the full list of restrictions please <a href="https://huggingface.co/spaces/CompVis/stable-diffusion-license" target="_blank" style="text-decoration: underline;" target="_blank">read the license</a></p>
-                    <p><h4>Biases and content acknowledgment</h4>
-Despite how impressive being able to turn text into image is, beware to the fact that this model may output content that reinforces or exacerbates societal biases, as well as realistic faces, pornography and violence. The model was trained on the <a href="https://laion.ai/blog/laion-5b/" style="text-decoration: underline;" target="_blank">LAION-5B dataset</a>, which scraped non-curated image-text-pairs from the internet (the exception being the removal of illegal content) and is meant for research purposes. You can read more in the <a href="https://huggingface.co/runwayml/stable-diffusion-v1-5" style="text-decoration: underline;" target="_blank">model card</a></p>
+                    Run AI models on your computer and avoid paying egregious cloud fees.
+                    This demo is powered by the <a href="https://faircompute.com/">FairCompute</a> platform.
                </div>
            """
         )

requirements.txt

@@ -1,4 +1,3 @@
 gradio < 4
-octoai-sdk
 faircompute==0.20.0
 httpx==0.24.1

text_to_image.py

@@ -1,17 +1,15 @@
 import base64
 import logging
 import os
+import requests
 import time
 from io import BytesIO
 
 from PIL import Image
-from octoai.client import Client as OctoAiClient
 from fair import FairClient
 
 logger = logging.getLogger()
 
-import requests
-
 SERVER_ADDRESS = "https://faircompute.com:8000"
 ENDPOINT_ADDRESS = "http://dikobraz.mooo.com:5000"
 TARGET_NODE = "119eccba-2388-43c1-bdb9-02133049604c"
@@ -19,19 +17,32 @@ TARGET_NODE = "119eccba-2388-43c1-bdb9-02133049604c"
 # ENDPOINT_ADDRESS = "http://localhost:5000"
 # TARGET_NODE = "ef09913249aa40ecba7d0097f7622855"
 
-DOCKER_IMAGE = "faircompute/diffusion-octo:v1"
+DOCKER_IMAGE = "faircompute/diffusers-api-sdxl-turbo"
 
 
 class EndpointClient:
-    def infer(self, prompt):
-        client = OctoAiClient()
+    def __init__(self, timeout):
+        response = requests.get(os.path.join(ENDPOINT_ADDRESS, 'healthcheck'), timeout=timeout).json()
+        if response['state'] != 'healthy':
+            raise Exception("Server is not healthy")
 
-        inputs = {"prompt": {"text": prompt}}
-        response = client.infer(endpoint_url=f"{ENDPOINT_ADDRESS}/infer", inputs=inputs)
-
-        image_b64 = response["output"]["image_b64"]
-        image_data = base64.b64decode(image_b64)
-        image_data = BytesIO(image_data)
+    def infer(self, prompt):
+        inputs = {
+            "modelInputs": {
+                "prompt": prompt,
+                "num_inference_steps": 4,
+                "guidance_scale": 0.0,
+                "width": 512,
+                "height": 512,
+            },
+            "callInputs": {
+                "MODEL_ID": "stabilityai/sdxl-turbo",
+                "PIPELINE": "AutoPipelineForText2Image",
+            },
+        }
+
+        response = requests.post(ENDPOINT_ADDRESS, json=inputs).json()
+        image_data = BytesIO(base64.b64decode(response["image_base64"]))
         image = Image.open(image_data)
 
         return image
@@ -41,18 +52,15 @@ class ServerNotReadyException(Exception):
     pass
 
 
-def wait_for_server(retries, timeout, delay=1.0):
+def wait_for_server(retries, timeout=1.0, delay=2.0):
     for i in range(retries):
         try:
-            r = requests.get(ENDPOINT_ADDRESS, timeout=timeout)
-            r.raise_for_status()
-            return
+            return EndpointClient(timeout=timeout)
         except (requests.exceptions.ConnectionError, requests.exceptions.HTTPError, requests.exceptions.Timeout) as e:
-            if i == retries - 1:
-                raise ServerNotReadyException("Failed to start the server") from e
-            else:
-                logger.info("Server is not ready yet")
-                time.sleep(delay)
+            logging.exception(e)
+            time.sleep(delay)
+
+    raise ServerNotReadyException("Failed to start the server")
 
 
 def start_server():
@@ -64,20 +72,17 @@ def start_server():
     client.run(node=TARGET_NODE,
                image=DOCKER_IMAGE,
                runtime="nvidia",
-               ports=[(5000, 8080)],
+               ports=[(5000, 8000)],
                detach=True)
 
-    # wait until the server is ready
-    wait_for_server(retries=10, timeout=1.0)
-
 
 def text_to_image(text):
     try:
-        wait_for_server(retries=1, timeout=1.0, delay=0.0)
+        client = wait_for_server(retries=1)
     except ServerNotReadyException:
         start_server()
+        client = wait_for_server(retries=10)
 
-    client = EndpointClient()
     return client.infer(text)