app.py

import streamlit as st
import os, io, ast #wget
import matplotlib.pyplot as plt
from PIL import Image
from genQC.pipeline.diffusion_pipeline import DiffusionPipeline
from genQC.inference.infer_srv import generate_srv_tensors, convert_tensors_to_srvs
from genQC.util import infer_torch_device

#--------------------------------
# download model into storage

#save_destination = "saves/"
#url_config  = "https://github.com/FlorianFuerrutter/genQC/blob/044f7da6ebe907bd796d3db293024db223cc1852/saves/qc_unet_config_SRV_3to8_qubit/config.yaml"
#url_weights = "https://github.com/FlorianFuerrutter/genQC/blob/044f7da6ebe907bd796d3db293024db223cc1852/saves/qc_unet_config_SRV_3to8_qubit/model.pt"

#def download(url, dst_dir):
#    if not os.path.exists(dst_dir): os.mkdir(dst_dir)
#    filename = os.path.join(dst_dir, os.path.basename(url))
#    if not os.path.exists(filename): filename = wget.download(url + "?raw=true", out=filename)
#    return filename

#config_file  = download(url_config, save_destination)
#weigths_file = download(url_weights, save_destination)

#--------------------------------
# setup

@st.cache_resource
def load_pipeline():
    #pipeline = DiffusionPipeline.from_config_file(save_destination, infer_torch_device())  
    pipeline = DiffusionPipeline.from_pretrained("Floki00/qc_srv_3to8qubit", "cpu")
    pipeline.scheduler.set_timesteps(20)   
    return pipeline

pipeline = load_pipeline()


is_gpu_busy = False
def get_qcs(srv, num_of_qubits, max_gates, g):
    global is_gpu_busy

    with st.status("Generation started", expanded=True) as status:
        st.write("Generating tensors...")
        out_tensor = generate_srv_tensors(pipeline, f"Generate SRV: {srv}", samples=6, system_size=num_of_qubits, num_of_qubits=num_of_qubits, max_gates=max_gates, g=g)
        
        st.write("Converting to circuits...")
        qc_list, _, srv_list = convert_tensors_to_srvs(out_tensor, pipeline.gate_pool)
        
        st.write("Plotting...")
        fig, axs = plt.subplots(3, 2, figsize=(7,10), constrained_layout=True, dpi=120)

        for ax in axs.flatten():
            ax.axis('off')
            ax.text(0.5, 0.5,"Circuit generated with errors")
            
        
        for qc,is_svr,ax in zip(qc_list, srv_list, axs.flatten()): 
            ax.clear()
            qc.draw("mpl", plot_barriers=False, ax=ax)
            ax.set_title(f"{'Correct' if is_svr==srv else 'NOT correct'}, is SRV = {is_svr}")
        status.update(label="Generation complete!", state="complete", expanded=False)
    
    # buf = io.BytesIO()
    # fig.savefig(buf)
    # buf.seek(0)
    # return Image.open(buf)
    return fig

#--------------------------------
# run

st.title("genQC · Generative Quantum Circuits")
st.write("""
Generating quantum circuits with diffusion models. Official demo of [[paper-arxiv]](https://arxiv.org/abs/2311.02041) [[code-repo]](https://github.com/FlorianFuerrutter/genQC).
""")

col1, col2 = st.columns(2)

srv           = col1.text_input('SRV', "[1,1,1,2,2,2]")
num_of_qubits = col1.radio('Number of qubits (should match SRV)', [3,4,5,6,7,8], index=3)
max_gates     = col1.select_slider('Max gates', options=[4,8,12,16,20,24,28], value=16)
g             = col1.slider('Guidance scale', min_value=0.0, max_value=15.0, value=10.0)

srv_list = ast.literal_eval(srv)
if len(srv_list)!=num_of_qubits:
    st.warning(f'Number of qubits does not match with given SRV {srv_list}. This could result in error-circuits!', icon="⚠️")

if col1.button('Generate circuits'):    
    fig = get_qcs(srv_list, num_of_qubits, max_gates, g)
    # col2.image(image, use_column_width=True)
    col2.pyplot(fig)