404 when executing `mlflow ui` from mlflow repos root folder about mlflow HOT 19 CLOSED

I found also this issue. Cloned the repo, followed documentation, mlruns folder is created in the repo, then I run mlflow ui, and my browser can't access it. Solution was to move mlruns folder out of the repo, then I could run mlfow ui and access it from the browser.

from mlflow.

Found the solution, you have got to change directory first :
The documentation is unclear, you just need to clone the examples not the whole repo.

cd example/tutorial/
pip install mlflow --user
python train.py
mlflow ui
# make sure you run this command in the directory where the /mlruns are

from mlflow.

I think your issue is the same as #49.

from mlflow.

I think it's the same as #108 actually, which is a an issue with Anaconda on Mac OS. The workaround there was to use http://localhost:5000 instead of http://127.0.0.1:5000. Does that fix the issue?

from mlflow.

Nope, I already reported in the issue that I was using http://localhost:5000.
Here's the which mlflow output:
/Users/arinto/anaconda3/envs/mlflow-exp/bin/mlflow

I'm using Conda environment in my mlflow installation.

from mlflow.

I tried to set my TCP/IP -> IPV6 setting as @dennyglee pointed out, but it still doesn't work. Let me try to restart my laptop.

from mlflow.

Restarting doesn't help. Will try to install mlflow to system python without anaconda

from mlflow.

from mlflow.

Unfortunately it still doesn't work (i.e. using MacOS python3, installed via Homebrew). Here's my config:

• Have I written custom code (as opposed to using a stock example script provided in MLflow): No
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Mac OS X 10.12.6 (Sierra)
• MLflow installed from (source or binary): binary
• MLflow version (run python -c "from mlflow import version; print(version.VERSION)"): 0.2.1
• Python version: Python 3.6.5
• which mlflow: /usr/local/bin/mlflow

Already set my TCP/IP-IPV6 to LinkLocalOnly

from mlflow.

Got it, let me see if I can test a couple of things out @arinto - and to confirm, this is not working for both http://127.0.0.1:5000 and http://localhost:5000?

from mlflow.

from mlflow.

I am facing the same issue. Install mlflow through pip3

from mlflow.

So, I was getting the same 404 error though I was able to run mlflow ui command. Here is some information about my system setup:

• I have a default anaconda3 installed.
• I receive ~/anaconda3/bin/mlflow and ~/anaconda3/bin/python3.6 paths upon executing both which mlflow and which python3.6
• I also have python2.7 installed which is not through anaconda. which python2.7 returns me /usr/bin/mlflow .

I copied the mlruns folder to /usr/bin/. Then executed mlflow ui command form /usr/bin$ .
My UI was displayed on both http://localhost:5000/ and http://127.0.0.1:5000/ .

from mlflow.

True. If mlflow is installed through pip3 install mlflow we can the UI at http://127.0.0.1:5000/.

from mlflow.

I ran into the same issue. The only way I can see the ui is that I create a new folder and move the mlruns into that folder and run mlflow ui

from mlflow.

Interestingly, it works fine with my new project(i.e. a simple Data Science project from scratch). Probably it was just a glitch.

from mlflow.

And it works fine with the sample project that I wrote in this issue.

Here's what I did:
As far as I remember, .bash_profile sets PATH environment variable with 2 Python executables (1 with Anaconda, 1 from with /Library/Frameworks/.... dir), and I removed the one with /Library/Frameworks/....

from mlflow.

I ran into this as well - mlflow ran for a two runs of examples/quickstart/test.py, one run of the scikit, then it stopped working on the second run of scikit (fourth run in total).

from mlflow.

Thanks all for the helpful suggestions! Running mlflow ui from within a checkout of MLflow runs the dev UI from source, which doesn't come pre-built with the JS/CSS assets needed to render the UI.

We recommend running mlflow ui from a directory other than your checkout of MLflow (the same workaround many of you already discussed above), using the --file-store option if needed to specify the path to the logging directory to view with the UI. #289 updates the quickstart/tutorial/README to clarify this :).

from mlflow.