uber-research / tailr

TAILR

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

TAILR

Run and enter the development container

You need Docker and make installed on your machine. From the project root run the following:

make build
make run-k8s-secrets

Run TAILR on local

Inside the container you can run

bazel run //src/tailr:train_main

Run all experiments of TAILR on GCloud

Inside the container you can run

bazel run //src/tailr:tailr2-gke

This will run all experiments described in the jobs variable of the doe_gke in the BUILD file. You can pick the experiments you want to run by changing the values of the jobs variable.

Visualize TB from inside a docker container

Enter the container with the command

make run-p<local_port>_<container_port>

This will create a link between the chosen local port and container port.

From the container now run

bazel run //src/tb_vis:main <log_dir>

to visualize the TensorBoard logs present in <log_dir>, where <log_dir> is an absolute path.

About

TAILR

License:Other


Languages

Language:Python 92.8%Language:Jupyter Notebook 3.5%Language:Starlark 2.6%Language:Shell 0.8%Language:Makefile 0.2%Language:Dockerfile 0.0%