Pipeline Power and Execution Limits: Unraveling This Week's Top Valohai Queries! 🚀

Question 1: Hey Team, I set up a task to adjust settings for better performance and logged data for 9 runs, including model settings and AUCROC scores. Now, I want to create a line chart to see how one setting (like max_depth) affects AUCROC across these runs. However, the chart isn't showing up. Am I doing something wrong?

Answer:
The thing is, we've only got one value per metadata per run, which makes impossible to whip up a line chart.

Combining the same metadata from different runs into one chart? Not doable, unfortunately.

Our plotting tools aren't super fancy – it's a conscious choice. We reckon there are better tools out there for this. 😊

Question 2: Hey there! Can I get the pipeline info from one of its node executions? Thanks!

Answer:
Absolutely! Here's how: In the UI: Simply click on the =<pipeline-id> in the execution table to access the pipeline. Alternatively, you can filter the executions table by the pipeline id. With API: Use the ExecutionRetrieve endpoint to fetch details about the pipeline the execution was involved in. https://app.valohai.com/api/v0/executions/<exec-id>/

Question 3: Hello! I have a question about passing the files in the pipeline:

I want to pass outputs from one node to another while also keeping all subdirectories, same behavior as keep-directories: suffix

Imagine I have output files like this in node 1 : valohai/outputs/site_A/test.json , valohai/outputs/site_B/test.json

I would like to define an edge and pass these outputs to node 2 inputs: • valohai/inputs/site_A/test.json • valohai/inputs/site_B/test.json Is that possible ?

Answer:
Yes, it is definitely possible.

You need to create an edge in your pipeline definition:

- [node1.output.*, node2.input.myinput]

Then node1 produces two outputs: /valohai/outputs/site_A/test.json /valohai/outputs/site_B/test.json

And the node2 gets these files as inputs: /valohai/inputs/myinput/site_A/test.json /valohai/inputs/myinput/site_A/test.json

Question 4: Can you help me with an issue? I'm having trouble fetching the latest commit for the main branch in my project. Fetching from other branches works fine. Could this be related to the transition from 'master' to 'main'? How can I fix this? Thanks!

Answer:
Of course! It's likely a simple fix. You just need to specify the branch you're fetching from in the UI. Go to project Settings → Repository → Make sure that the branches are all set up.

Once you do that, you should be all set! 😊

Question 5: 🚀 Hey there! What's the maximum number of parallel executions we can currently run on Valohai? Is there a limit at all?

Answer:
Great question!

While Valohai doesn't impose a direct limit on parallel executions, there might be AWS-related quotas based on machine types or GPUs.

Additionally, AWS availability could impact the number of machines you can utilize simultaneously. If you encounter execution delays, it's wise to check your environment settings for any relevant messages.

Also, ensure that the max queued executions for your Task is set to an appropriate number. The default is 1000, which should suffice for most scenarios. Happy coding! 🌟