scalene VS Task

I collected a list of profilers (also memory profilers, also specifically for Python) here: https://github.com/albertz/wiki/blob/master/profiling.md

Currently I actually need a Python memory profiler, because I want to figure out whether there is some memory leak in my application (PyTorch based training script), and where exactly (in this case, it's not a problem of GPU memory, but CPU memory).

I tried Scalene (https://github.com/plasma-umass/scalene), which seems to be powerful, but somehow the output it gives me is not useful at all? It doesn't really give me a flamegraph, or a list of the top lines with memory allocations, but instead it gives me a listing of all source code lines, and prints some (very sparse) information on each line. So I need to search through that listing now by hand to find the spots? Maybe I just don't know how to use it properly.

I tried Memray, but first ran into an issue (https://github.com/bloomberg/memray/issues/212), but after using some workaround, it worked now. I get a flamegraph out, but it doesn't really seem accurate? After a while, there don't seem to be any new memory allocations at all anymore, and I don't quite trust that this is correct.

There is also Austin (https://github.com/P403n1x87/austin), which I also wanted to try (have not yet).

Somehow this experience so far was very disappointing.

(Side node, I debugged some very strange memory allocation behavior of Python before, where all local variables were kept around after an exception, even though I made sure there is no reference anymore to the exception object, to the traceback, etc, and I even called frame.clear() for all frames to really clear it. It turns out, frame.f_locals will create another copy of all the local variables, and the exception object and all the locals in the other frame still stay alive until you access frame.f_locals again. At that point, it will sync the f_locals again with the real (fast) locals, and then it can finally free everything. It was quite annoying to find the source of this problem and to find workarounds for it. https://github.com/python/cpython/issues/113939)

Use this my fren: https://github.com/plasma-umass/scalene

You should take a look at Scalene - it's even better.

https://github.com/plasma-umass/scalene

I like seeing another Python profiler. The one I've been playing with is Scalene (GitHub). It does some fun things related to letting you see how much things are moving across the system Python memory boundary.

Ai vazut "Python Performance Matters" by Emery Berger (Strange Loop 2022)? E in principiu o prezentare si demo cu Scalene.

I tried out some features of machine learning models suggesting optimisations on code profiled by scalene and pretty much all of them would make the code less efficient, both time and memory wise. I am not worried. The devil is in the details and ML will not replace all of us anytime soon

So many tools in this space! This one looks a little bit like go-task, but it seems maybe better for production workflows because if timeout support, while go-task seems more aimed to command line work/makefile replacement.

—-

https://github.com/go-task/task

View on GitHub

A similar tool is `task` https://taskfile.dev/ . It is quite capable and also a single executable. I've grown to quite like it.

check out tasks - a bit of a learning curve but arguably more powerful imo

That's when I came across taskfile.dev. Task is an automation tool designed to be more accessible than other options, such as GNU Make.

Full disclosure, I did not review Conda or Hatch fully. Not that there is anything explicitly wrong with either of them. Conda is too specific to the scientific community for my general taste. Hatch seems to go well with Conda and also uses the PyProject manifest as well. It's nice that it gives you several built in tools, similar to commit hooks, but I tend to like to roll my own via a Taskfile and run them with Poetry.

Taskfile is a tool for streamlining repetitive development tasks. It helps automate activities like building, testing, and deploying applications. Unlike Makefile, Taskfile uses YAML for configuration, making it more readable and user-friendly.

9. We test everything with another promotion which runs make targets which build docker containers to run python scripts (pytest)

This is also built by a complicated web of wildcarded makefile targets, which need to be interoperable and support a few if/else cases for specific components.

My plan is to migrate all of this to something simpler and more straightforward, or at least more maintainable, which is honestly probably going to turn into taskfile[0] instead of makefiles, and then simple python scripts for the glue that ties everything together or does more complex logic.

My hope is that it can be more straightforward and easier to maintain, with more component-ized logic, but realistically every step in that labyrinthine build process (and that's just the open-source version!) came from a decision made by a very talented team of engineers who know far more about the process and the product than I do. At this point I'm wondering if it would make 'more sense' to replace it with a giant python script of some kind and get access to all the logic we need all at once (it would not).

[0] https://taskfile.dev/

task - a task runner and a replacement for make