http://projector.tensorflow.org/ allows you to publish and share your own visualizations if you click "publish" and link to your own .json config gist w/ path to your own data. However, I'm getting an error when I link it to my gist .json config file as the instructions dictate (https://gist.github.com/jmolayem/2a28b68807602e4c3c8fa9987c1d08a6):
{
"embeddings": [
{
"tensorName": "My tensor",
"tensorShape": [
1000,
50
],
"tensorPath": "https://gist.github.com/jmolayem/2a28b68807602e4c3c8fa9987c1d08a6#file-test55-tsv"
}
]
}
My hunch is the tensorShape has to match the .tsv file. But if that were true, I don't understand why the direct upload of the .tsv file works in the demo but not when I publish.
Am I missing something in formatting the gist file?
You are pretty close! Your tensorPath
doesn't point to the raw tsv file, but to an html page showing a snippet of that file, along other files.
If you click on the raw
button on the top left of the test55.tsv
snippet, you will get
https://gist.githubusercontent.com/jmolayem/2a28b68807602e4c3c8fa9987c1d08a6/raw/15252f862a318b8b11e550bdf2bfddb3991f4cab/test55.tsv
which you should use as a tensorPath
.