Add ability to select individual files from job output

[mrahtz]

My workflow involves setting off multiple runs with different parameters then examining TensorBoard results from each run. At the moment, this involves browsing to each job's output directory on the web interface, finding the TensorBoard events files, downloading, and renaming them to match the job ID manually. (I don't want to just do a floyd data clone because jobs produce large outputs, and I usually only need the events files.) It would be great if there could be a way to automate this.

Ideally it would be something like:

floyd data listfiles fooproject/1/output
<parse output to find the name of the events file>
floyd data getfile fooproject/1/output tensorboard/events-18384

From a quick look at the floyd data code, though, I wasn't sure what API calls this would involve. Is there an existing API call for listing files in a dataset/downloading individual files?

Thanks! 🙂

[mrahtz]

Ah, never mind. Figured it out from the web interface.