routers/training.py

from fastapi import APIRouter, Form, BackgroundTasks
from config import settings
import os
import json
import utils
import torch
import requests
from PIL import Image
from io import BytesIO
from pydantic import BaseModel
import base64
import uuid
from transformers import AutoTokenizer, AutoModelForCausalLM

from diffusers import StableDiffusionImg2ImgPipeline
from diffusers import StableDiffusionInpaintPipeline

# tokenizer = AutoTokenizer.from_pretrained("openlm-research/open_llama_7b")
# model = AutoModelForCausalLM.from_pretrained(
#     "openlm-research/open_llama_7b", device_map="auto", load_in_4bit=True
# )

# model_id_or_path = "runwayml/stable-diffusion-v1-5"
# pipe = StableDiffusionImg2ImgPipeline.from_pretrained(model_id_or_path, torch_dtype=torch.float16)


pipe = StableDiffusionInpaintPipeline.from_pretrained(
    "runwayml/stable-diffusion-inpainting",
    revision="fp16",
    torch_dtype=torch.float16,
)

pipe = pipe.to("cuda")


router = APIRouter()

class ActionBody(BaseModel):
    url: str
    maskUrl: str
    prompt: str
    strength: float
    guidance_scale: float
    resizeW: int
    resizeH: int

@router.post("/perform-action")
async def performAction(actionBody: ActionBody):
    
    # model_inputs = tokenizer(["A list of colors: red, blue"], return_tensors="pt").to("cuda")
    # generated_ids = model.generate(**model_inputs)
    # output = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

    response = requests.get(actionBody.url)
    init_image = Image.open(BytesIO(response.content)).convert("RGB")
    init_image = init_image.resize((actionBody.resizeW, actionBody.resizeH))

    response = requests.get(actionBody.maskUrl)
    mask_image = Image.open(BytesIO(response.content)).convert("RGB")
    mask_image = mask_image.resize((actionBody.resizeW, actionBody.resizeH))

    
    # images = pipe(prompt=actionBody.prompt, image=init_image,  strength=actionBody.strength, guidance_scale=actionBody.guidance_scale).images
    images = pipe(prompt=actionBody.prompt, image=init_image,  mask_image=mask_image).images
    print(images)
    buffered = BytesIO()
    images[0].save(buffered, format="JPEG")
    img_str = base64.b64encode(buffered.getvalue())
    imgUUID = str(uuid.uuid4())
    images[0].save(imgUUID+".png")

    return {
        "imageName" : imgUUID+".png",
        "image": "data:image/jpeg;base64,"+img_str.decode(),
        # "output": output
    }


@router.get("/hi")
async def hifunction():
    
    url = "https://raw.githubusercontent.com/CompVis/stable-diffusion/main/assets/stable-samples/img2img/sketch-mountains-input.jpg"
    response = requests.get(url)
    init_image = Image.open(BytesIO(response.content)).convert("RGB")
    init_image = init_image.resize((768, 512))
    prompt = "A fantasy landscape, trending on artstation"
    images = pipe(prompt=prompt, image=init_image, strength=0.75, guidance_scale=7.5).images
    print(images)
    print(images[0])

    buffered = BytesIO()
    images[0].save(buffered, format="JPEG")
    img_str = base64.b64encode(buffered.getvalue())

    # images[0].save("fantasy_landscape.png")

    return {
        "image": "data:image/jpeg;base64,"+img_str.decode()
    }