app.py

import numpy as np
import streamlit as st
import torch
import matplotlib.pyplot as plt

import disvae
import transforms as trans

P_MODEL = "model/drilling_ds_btcvae"

@st.cache_resource
def load_decode_function():
    sorter = trans.LatentSorter(disvae.get_kl_dict(P_MODEL))
    vae = disvae.load_model(P_MODEL)
    scaler = trans.MinMaxScaler(_min=torch.tensor([1.3]),_max=torch.tensor([4.0]),min_norm=0.3,max_norm=0.6)
    imaging = trans.SumField()

    _dec = trans.sequential_function(
        sorter.inv,
        vae.decoder,
        scaler.inv,
        imaging.inv
    )

    def decode(latent):
        with torch.no_grad():
            return trans.np_sample(_dec)(latent)
    
    return decode
    
decode = load_decode_function()

st.markdown("**Latent Space Parameters**")
latent_vector = np.array([st.slider(f"Latent Dimension {l}",min_value=-3.0,max_value=3.0,value=0.0) for l in range(3)])
latent_vector = np.concatenate([latent_vector,np.zeros(7)],axis=0)

ts = decode(latent_vector)

st.markdown("**Generated Time Series**")

fig, ax = plt.subplots(figsize=(8,4))

ax.plot(ts.ravel())
ax.set_ylim([0,4])

st.pyplot(fig)