K3D on Google Colab

[effepivi]

• K3D version: 2.15.2
• Python version: 3.9.16 (main, Dec 7 2022, 01:11:51) [GCC 9.4.0]
• Operating System: Linux

Description

I am trying to use K3D on Google Colab. The viewer appears, I can interact with it (the grid moves) but the list of objects remains empty. I have tested using my own code, and also k3D examples. The issue always happens. See below for a screenshot and the corresponding code.

What I Did

I recycled one of K3D's examples:

!pip install k3d --quiet

from google.colab import output
output.enable_custom_widget_manager()

import sys
import platform
import k3d
import random
import numpy as np
import matplotlib

print("Platform:", platform.system())
print("K3D:", k3d.__version__)
print("Python:", sys.version)

T = 1500
N = 10
P = 100
mod = 50

positions = [None] * N
colors = [None] * N
trajectories = [None] * N
colormaps = [k3d.colormaps.matplotlib_color_maps.hot]

for n in range(N):
    positions[n] = {}
    trajectories[n] = {}
    bang = False

    pos = np.zeros((P, 3), dtype=np.float32)
    velocity = np.zeros((P, 3), dtype=np.float32)
    velocity[:, 2] = 180.0 + np.random.random(P) * 0.1    

    pos[:, 0] = np.sin(n/N * 2.0 * np.pi) * 10.0
    pos[:, 1] = np.cos(n/N * 2.0 * np.pi) * 10.0

    cm = random.choice(colormaps)
    cm_space = np.linspace(np.min(cm[0::4]), np.max(cm[0::4]), 2048)

    c = (np.interp(cm_space, cm[::4], cm[3::4]) * 255).astype(np.uint32) + \
        (np.interp(cm_space, cm[::4], cm[2::4]) * 255).astype(np.uint32) * 256 + \
        (np.interp(cm_space, cm[::4], cm[1::4]) * 255).astype(np.uint32) * 256 ** 2

    colors[n] = c[np.random.randint(0, c.shape[0], P)]

    for t in range(T):
        if t > n * 30:        
            if np.mean(velocity[:, 2] < 30.0) and not bang:
                bang = True
                velocity[:, 0] += np.random.normal(size=P) * 55
                velocity[:, 1] += np.random.normal(size=P) * 55
                velocity[:, 2] += np.random.normal(size=P) * 35

            velocity[:, 2] -= 0.05
            velocity *= 0.995                
            pos += velocity * 0.001

        if t % mod == 0:
            positions[n][str(t/100.0)] = pos.copy()

for n in range(N):
    db = np.array([positions[n][k] for k in positions[n].keys()])

    for t in range(T//mod):
        anim = np.empty((P, T//mod + 1, 3), dtype=np.float32)
        anim.fill(np.nan)
        for p in range(P):
            anim[p, 0:t+1, :] = db[0:t+1, p]
            anim[p, t + 1, :] = anim[p, t, :]

        trajectories[n][str(t * mod/100.0)] = anim

plot = k3d.plot(grid=[-100,-100,-100,100,100,100], grid_auto_fit=True, camera_auto_fit=True)

for n in range(N):
    plot += k3d.points(positions[n], colors=colors[n], shader='3d', opacity=0.85, point_size=0.5)
    plot += k3d.line(trajectories[n], shader='simple')

text_positions= {
    '0.0': np.array([0,0,100]),
    '4.3': np.array([0,0,100]),
    '4.8': np.array([0,0,20])
}

text_size= {
    '0.0': 1,
    '4.0': 1,
    '4.8': 80
}

plot += k3d.texture_text('Happy New Year!', text_positions, size=text_size, color=0xaaaaaa)
plot.display()

Web console log / python logs

/usr/local/lib/python3.9/dist-packages/traittypes/traittypes.py:97: UserWarning: Given trait value dtype "int64" does not match required type "float32". A coerced copy has been created.

In case of web console log please start from point like:

K3D: (UNMASKED_VENDOR_WEBGL) Google Inc. (NVIDIA)
...

up to the last line. If you don't want to share details about your GPU please anonymise it

[Mr-McGL]

Hi,

A similar issue is reported in this thread [X].

So far, you can display static plots using inline snapshots [X].

%%capture __pip_output__
!pip install k3d
!pip install pyvista

import numpy as np
import k3d
from pyvista import examples

from IPython.display import HTML

You don't need to enable custom widgets

# from google.colab import output
# output.enable_custom_widget_manager()

# https://k3d-jupyter.org/gallery/showcase/terrain.html
dem = examples.download_crater_topo()
data = dem.get_array(0).reshape(dem.dimensions[::-1])[0, :, :].astype(np.float32)

plot = k3d.plot()

obj = k3d.surface(data,
            attribute=data,
            flat_shading=False,
            color_map = k3d.colormaps.matplotlib_color_maps.viridis,
            xmin=dem.bounds[0],
            xmax=dem.bounds[1],
            ymin=dem.bounds[2],
            ymax=dem.bounds[3])

plot += obj
#plot.display()

plot.snapshot_type = 'inline'
HTML(plot.get_snapshot())

--or--

plot.snapshot_type = 'inline'
with open('output.html', 'w') as f:
f.write(plot.get_snapshot())

HTML(filename='output.html')

[effepivi]

Thanks, it works!

[artur-trzesiok]

Good news @effepivi , @Mr-McGL !

Experimental COLAB support is done: https://colab.research.google.com/drive/1aZd9vvPJBZzlX5bmaAIp8eE9f_0c3IaS?usp=sharing

[effepivi]

@artur-trzesiok

Thanks for letting me know. I'll try it!

[Mr-McGL]

Thanks. It works!!!