tensorboard icon indicating copy to clipboard operation
tensorboard copied to clipboard

load data from relative path with the projector plugin

Open teamdandelion opened this issue 8 years ago • 3 comments

Migrated from https://github.com/tensorflow/tensorflow/issues/8342

This is a problem similar with issue #7382 but it has been closed.

@dandelionmane Another place that still has the similar problem in #7382 is the request of project plugin data in Embeddings panel, e.g. /data/plugin/projector/runs still requests in a absolute path. I found the code here , there is a leading slash for the route-prefix property in vz-projector-dashboard. If that is correct, I will submit a PR. Thanks!

Another question is when will the tensorboard be recompiled. As now in version 1.0 these features are not included.

teamdandelion avatar Jun 16 '17 18:06 teamdandelion

I'm also experiencing this issue. The scenario is simple:

  1. Multiple cloud instances are writing Tensorboard embedding logs to a folder that is mounted as /output to each of the instances.
  2. The actual folder for each instance is located at /output/uuid on a central server.
  3. Tensorboard is initialized such that log-dir is /output on the central server.
  4. Scalar summaries show all runs as separate runs from the different uuid folders. 😄
  5. Embedding projector does not find the meta file and is stuck after receiving a 400 error.

Once the projector finds the run files, why not go up the tree to find the meta file instead of starting at the root, which is obviously different for Tensorboard and for the Saver, and going down the folder tree?

ofirbb avatar Jun 18 '17 14:06 ofirbb

Not sure if this is related to this issue or the previous issue. But I found that in r1.0, if I am in the $ROOT directory and write logs to $ROOT/$LOG_DIR directory with $ROOT/$LOG_DIR/metadata.tsv and $ROOT/$LOG_DIR/sprite.png. I could do:

config = projector.ProjectorConfig()
embed = config.embeddings.add()
embed.tensor_name = embedding.name

embed.metadata_path = '$LOG_DIR/metadata.tsv'
embed.sprite.image_path = '$LOG_DIR/sprite.png'

But after r1.1, I should do:

config = projector.ProjectorConfig()
embed = config.embeddings.add()
embed.tensor_name = embedding.name

embed.metadata_path = 'metadata.tsv'
embed.sprite.image_path = 'sprite.png'

I am putting this note here in case some people also find that the embeddings tab suddenly stops to work after upgrading tf, and eventually find this issue page.

appierys avatar Jul 04 '17 04:07 appierys

Is this issue resolved ?

Harshini-Gadige avatar Mar 05 '19 23:03 Harshini-Gadige