Skip to content

Instantly share code, notes, and snippets.

@merutak
Created March 21, 2021 08:01
Show Gist options
  • Save merutak/9432e0cfeca8f4cfc9fd6210fa1081ef to your computer and use it in GitHub Desktop.
Save merutak/9432e0cfeca8f4cfc9fd6210fa1081ef to your computer and use it in GitHub Desktop.
import numpy matrix to display in tensorboard
import os
from tensorboard.plugins import projector
import math
from django.conf import settings
log_dir='/logs/abcd/'
if not os.path.exists(log_dir):
os.makedirs(log_dir)
# Save Labels separately on a line-by-line manner.
with open(os.path.join(log_dir, 'metadata.tsv'), "w") as f:
f.write(f"name\tCategory\n")
for name, category in my_point_names: 10))
f.write(f"{name}\t{category}\n")
weights = tf.Variable(embedding) # assumed to be in the same order as 'my_point_names'
checkpoint = tf.train.Checkpoint(embedding=weights)
checkpoint.save(os.path.join(log_dir, f"embedding.ckpt"))
config = projector.ProjectorConfig()
embedding = config.embeddings.add()
embedding.tensor_name = f"embedding/.ATTRIBUTES/VARIABLE_VALUE"
embedding.metadata_path = 'metadata.tsv'
projector.visualize_embeddings(log_dir, config)
%tensorboard --logdir /logs/abcd/
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment