seeing if runtime error can be fixed

app.py

@@ -8,8 +8,15 @@ import numpy as np
 import numpy.typing as npt
 from dcgan import DCGAN3D_G
 import os
+import pathlib
 pv.start_xvfb()
 
+STREAMLIT_STATIC_PATH = pathlib.Path(st.__path__[0]) / 'static'
+
+DOWNLOADS_PATH = (STREAMLIT_STATIC_PATH / "downloads")
+if not DOWNLOADS_PATH.is_dir():
+    DOWNLOADS_PATH.mkdir()
+
 def download_checkpoint(url: str, path: str) -> None:
     resp = requests.get(url)
 
@@ -61,87 +68,91 @@ def create_matplotlib_figure(img: npt.ArrayLike, midpoint: int):
         a.set_axis_off()
     return fig
 
-st.title("Generating Porous Media with GANs")
-
-st.markdown(
-    """
-    ### Author
-    _[Lukas Mosser](https://scholar.google.com/citations?user=y0R9snMAAAAJ&hl=en&oi=ao) (2022)_ - :bird:[porestar](https://twitter.com/porestar)
-
-    ## Description
-    This is a demo of the Generative Adversarial Network (GAN, [Goodfellow 2014](https://arxiv.org/abs/1406.2661)) trained for our publication [PorousMediaGAN](https://github.com/LukasMosser/PorousMediaGan)
-    published in Physical Review E ([Mosser et. al 2017](https://journals.aps.org/pre/abstract/10.1103/PhysRevE.96.043309))
+def main():
+    st.title("Generating Porous Media with GANs")
 
-    The model is a pretrained 3D Deep Convolutional GAN ([Radford 2015](https://arxiv.org/abs/1511.06434)) that generates a volumetric image of a porous medium, here a Berea sandstone, from a set of pretrained weights.  
-    
-    ## Intent
-    I hope this encourages others to create interactive demos of their research for knowledge sharing and validation.
-    
-    ## The Demo
-    Slices through the 3D volume are rendered using [PyVista](https://www.pyvista.org/) and [PyThreeJS](https://pythreejs.readthedocs.io/en/stable/)
-    
-    The model itself currently runs on the :hugging_face: [Huggingface Spaces](https://huggingface.co/spaces) instance.  
-    Future migration to the :hugging_face: [Huggingface Models](https://huggingface.co/models) repository is possible.
+    st.markdown(
+        """
+        ### Author
+        _[Lukas Mosser](https://scholar.google.com/citations?user=y0R9snMAAAAJ&hl=en&oi=ao) (2022)_ - :bird:[porestar](https://twitter.com/porestar)
     
-    ### Interactive Model Parameters
-    The GAN used here in this study is fully convolutional "_Look Ma' no MLP's_": Changing the spatial extent of the latent space vector _z_
-    allows one to generate larger synthetic images. 
+        ## Description
+        This is a demo of the Generative Adversarial Network (GAN, [Goodfellow 2014](https://arxiv.org/abs/1406.2661)) trained for our publication [PorousMediaGAN](https://github.com/LukasMosser/PorousMediaGan)
+        published in Physical Review E ([Mosser et. al 2017](https://journals.aps.org/pre/abstract/10.1103/PhysRevE.96.043309))
     
-    """
-, unsafe_allow_html=True)
-
-view_width = 400
-view_height = 400
-
-model_fname = "berea_generator_epoch_24.pth"
-checkpoint_url = "https://github.com/LukasMosser/PorousMediaGan/blob/master/checkpoints/berea/{0:}?raw=true".format(model_fname)
-
-download_checkpoint(checkpoint_url, model_fname)
-
-latent_size = st.slider("Latent Space Size z", min_value=1, max_value=5, step=1)
-img = generate_image(model_fname, latent_size=latent_size)
-slices, mesh, dist = create_uniform_mesh_marching_cubes(img)
-
-pv.set_plot_theme("document")
-pl = pv.Plotter(shape=(1, 1),
-                     window_size=(view_width, view_height))
-_ = pl.add_mesh(slices, cmap="gray")
-pl.export_html('slices.html')
-
-pl = pv.Plotter(shape=(1, 1),
-                     window_size=(view_width, view_height))
-_ = pl.add_mesh(mesh, scalars=dist)
-pl.export_html('mesh.html')
-
-st.header("2D Cross-Section of Generated Volume")
-fig = create_matplotlib_figure(img, img.shape[0]//2)
-st.pyplot(fig=fig)
-
-
-
-HtmlFile = open("slices.html", 'r', encoding='utf-8')
-source_code = HtmlFile.read()
-st.header("3D Intersections")
-components.html(source_code, width=view_width, height=view_height)
-st.markdown("_Click and drag to spin, right click to shift._")
-
-HtmlFile = open("mesh.html", 'r', encoding='utf-8')
-source_code = HtmlFile.read()
-st.header("3D Pore Space Mesh")
-components.html(source_code, width=view_width, height=view_height)
-st.markdown("_Click and drag to spin, right click to shift._")
-
-st.markdown("""
-    ## Citation
-    If you use our code for your own research, we would be grateful if you cite our publication:
-    ```
-    @article{pmgan2017,
-        title={Reconstruction of three-dimensional porous media using generative adversarial neural networks},
-        author={Mosser, Lukas and Dubrule, Olivier and Blunt, Martin J.},
-        journal={arXiv preprint arXiv:1704.03225},
-        year={2017}
-    }```
-    """)
-
-os.remove("slices.html")
-os.remove("mesh.html")
+        The model is a pretrained 3D Deep Convolutional GAN ([Radford 2015](https://arxiv.org/abs/1511.06434)) that generates a volumetric image of a porous medium, here a Berea sandstone, from a set of pretrained weights.  
+        
+        ## Intent
+        I hope this encourages others to create interactive demos of their research for knowledge sharing and validation.
+        
+        ## The Demo
+        Slices through the 3D volume are rendered using [PyVista](https://www.pyvista.org/) and [PyThreeJS](https://pythreejs.readthedocs.io/en/stable/)
+        
+        The model itself currently runs on the :hugging_face: [Huggingface Spaces](https://huggingface.co/spaces) instance.  
+        Future migration to the :hugging_face: [Huggingface Models](https://huggingface.co/models) repository is possible.
+        
+        ### Interactive Model Parameters
+        The GAN used here in this study is fully convolutional "_Look Ma' no MLP's_": Changing the spatial extent of the latent space vector _z_
+        allows one to generate larger synthetic images. 
+        
+        """
+    , unsafe_allow_html=True)
+
+    view_width = 400
+    view_height = 400
+
+    model_fname = "berea_generator_epoch_24.pth"
+    checkpoint_url = "https://github.com/LukasMosser/PorousMediaGan/blob/master/checkpoints/berea/{0:}?raw=true".format(model_fname)
+
+    download_checkpoint(checkpoint_url, (DOWNLOADS_PATH / model_fname))
+
+    latent_size = st.slider("Latent Space Size z", min_value=1, max_value=5, step=1)
+    img = generate_image((DOWNLOADS_PATH / model_fname), latent_size=latent_size)
+    slices, mesh, dist = create_uniform_mesh_marching_cubes(img)
+
+    pv.set_plot_theme("document")
+    pl = pv.Plotter(shape=(1, 1),
+                         window_size=(view_width, view_height))
+    _ = pl.add_mesh(slices, cmap="gray")
+    pl.export_html((DOWNLOADS_PATH / 'slices.html'))
+
+    pl = pv.Plotter(shape=(1, 1),
+                         window_size=(view_width, view_height))
+    _ = pl.add_mesh(mesh, scalars=dist)
+    pl.export_html((DOWNLOADS_PATH / 'mesh.html'))
+
+    st.header("2D Cross-Section of Generated Volume")
+    fig = create_matplotlib_figure(img, img.shape[0]//2)
+    st.pyplot(fig=fig)
+
+
+
+    HtmlFile = open((DOWNLOADS_PATH / 'slices.html'), 'r', encoding='utf-8')
+    source_code = HtmlFile.read()
+    st.header("3D Intersections")
+    components.html(source_code, width=view_width, height=view_height)
+    st.markdown("_Click and drag to spin, right click to shift._")
+
+    HtmlFile = open((DOWNLOADS_PATH / 'mesh.html'), 'r', encoding='utf-8')
+    source_code = HtmlFile.read()
+    st.header("3D Pore Space Mesh")
+    components.html(source_code, width=view_width, height=view_height)
+    st.markdown("_Click and drag to spin, right click to shift._")
+
+    st.markdown("""
+        ## Citation
+        If you use our code for your own research, we would be grateful if you cite our publication:
+        ```
+        @article{pmgan2017,
+            title={Reconstruction of three-dimensional porous media using generative adversarial neural networks},
+            author={Mosser, Lukas and Dubrule, Olivier and Blunt, Martin J.},
+            journal={arXiv preprint arXiv:1704.03225},
+            year={2017}
+        }```
+        """)
+
+    #os.remove("slices.html")
+    #os.remove("mesh.html")
+
+if __name__ == "__main__":
+    main()