initial commit

Dockerfile.hf

@@ -0,0 +1,28 @@
+# Use the official Python 3.9 image
+FROM python:3.9
+
+# Set the working directory to /code
+WORKDIR /code
+
+# Copy the current directory contents into the container at /code
+COPY ./requirements.txt /code/requirements.txt
+
+# Install requirements.txt
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+
+# Set up a new user named "user" with user ID 1000
+RUN useradd -m -u 1000 user
+# Switch to the "user" user
+USER user
+# Set home to the user's home directory
+ENV HOME=/home/user \\
+    PATH=/home/user/.local/bin:$PATH
+
+# Set the working directory to the user's home directory
+WORKDIR $HOME/app
+
+# Copy the current directory contents into the container at $HOME/app setting the owner to the user
+COPY --chown=user . $HOME/app
+
+# Start the FastAPI app on port 7860, the default port expected by Spaces
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,10 +1,41 @@
 ---
-title: Sd Xl Refiner 1.0
+#Stable Diffusion Xl Refiner 1.0
+
+Our goal is to generate high quality fashion items (with models) based on the prompt.
+
+Steps (POC):
+- submit prompt on the UI to our AWS Lambda endpoint
+- take the prompt and feed to an LLM model to check if it's fashion related prompt.
+- If no, return an error.
+- if yes,
+  - increment distributed cache by 1. Return error when reaching 10 requests per hour. 
+  - feed the prompt to the HuggingFace Spaces endpoint
+- generate image, return image to the lambda 
+- save image in S3.
+- return image to the front-end
+- Any errors shown to the user, also show a subscription form to notify the user when we go live.
+
+What I need:
+- Serverless Framework
+- Fast API service unik-ml
+- Fast API service unik-huggingface-sd
+- S3 bucket
+- smallest distributed cache
+- Pulumi:
+  - Redis
+  - s3 bucket
+  - 
+
 emoji: 📈
+
 colorFrom: pink
+
 colorTo: gray
+
 sdk: docker
+
 pinned: false
+
 license: openrail
 ---
 

main.py

@@ -0,0 +1,64 @@
+from io import BytesIO
+
+import torch
+from diffusers import DiffusionPipeline
+from fastapi import FastAPI
+from fastapi.responses import StreamingResponse
+
+# load both base & refiner
+base = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True
+)
+base.to("cuda")
+# base.enable_model_cpu_offload()
+base.enable_attention_slicing()
+refiner = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-xl-refiner-1.0",
+    text_encoder_2=base.text_encoder_2,
+    vae=base.vae,
+    torch_dtype=torch.float16,
+    use_safetensors=True,
+    variant="fp16",
+)
+refiner.to("cuda")
+# refiner.enable_model_cpu_offload()
+refiner.enable_attention_slicing()
+
+# Create a new FastAPI app instance
+app = FastAPI()
+
+
+# Define a function to handle the GET request at `/generate`
+# The generate() function is defined as a FastAPI route that takes a
+# string parameter called text. The function generates text based on the # input using the pipeline() object, and returns a JSON response
+# containing the generated text under the key "output"
+@app.get("/generate")
+def generate(text: str):
+    """
+    generate image
+    """
+    # Define how many steps and what % of steps to be run on each experts (80/20) here
+    n_steps = 40
+    high_noise_frac = 0.8
+    negative = "drawing, painting, crayon, sketch, graphite, impressionist, noisy, blurry, soft, deformed, ugly. bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, disgusting, poorly drawn hands, missing limb, floating limbs, disconnected limbs, malformed hands, blurry, mutated hands and fingers, watermark, watermarked, oversaturated, censored, distorted hands, amputation, missing hands, obese, doubled face, double hands, two women, anime style, cartoon, toon."
+    prompt = "Designs should play with different textures and layering but stick to a monochrome palette. Think leather jackets over mesh tops, or satin draped over matte cotton. in a studio. zoomed-in. single model."
+
+    # run both experts
+    image = base(
+        prompt=prompt,
+        negative_prompt=negative,
+        num_inference_steps=n_steps,
+        denoising_end=high_noise_frac,
+        output_type="latent",
+    ).images
+    final_image = refiner(
+        prompt=prompt,
+        negative_prompt=negative,
+        num_inference_steps=n_steps,
+        denoising_start=high_noise_frac,
+        image=image,
+    ).images[0]
+
+    return StreamingResponse(BytesIO(final_image), media_type="image/png")
+    # Return the generated text in a JSON response
+    # return {"output": output[0]["generated_text"]}

requirements.txt

@@ -0,0 +1,7 @@
+fastapi==0.100.1
+pydantic==2.1.1
+pylint==2.17.5
+uvicorn>=0.23.2
+torch==2.0.1
+transformers==4.31.0
+accelerate==0.21.0