Unreliability of GIF Animating HTML in

Short Course Q&A How Diffusion Models Work
1

There is an unreliability (sometimes works and sometimes doesn’t with no changes), of the “gif animating” in Visualize Samples in Neural Network (L1_Sample, Cells 9 and 11) and Speeding Up (Lab 4, Fast Sampling, Cell 9 and 12).

Attached is example Video of Cells 22, 23 and 24 where it seems to unconnect to the Kernel in latter Cell 9 erroring out at gif animating frame 24 of 25, which reads as follows in Cell 9 of Speeding Up,

# visualize samples
plt.clf()
samples, intermediate = sample_ddim(32, n=25)
animation_ddim = plot_sample(intermediate,32,4,save_dir, "ani_run", None, save=False)
HTML(animation_ddim.to_jshtml())

Note how at the end of the video below it has a brown error message saying “Connecting to Kernel” and then reconnects to Kernel. Is this just a load balance issue/error we have to keep trying until it’s successful?
Google Drive Video

NOTE: I noticed that the “Neural Network” section loads all of its libraries in Cell 1 without error, whereas the “Speeding Up” section errors out see photos.

Neural Network

Neural Networks Cell 1
Neural Networks Cell 11334×1220 128 KB

Speeding Up

Speeding Up Cell 1 (error)
Speeding Up Cell 1 (error)1314×1246 134 KB

Please advise. Thanks,
Robert Thompson

2

It appears when executing the statement,

HTML(animation_ddpm_context.to_jshtml())

That you need to open the notebook in a separate tab. To do this Shift+Click the paragraph ❡ character in the first cell. The 3 window format is somehow ‘off’ for the HTML insert to work.