No worries Andriy,
Still masses better than the Flame one and saved me a heap of pain on the job.
Hope you’re OK
I’m having problems installing this on a mac, (non M1 running mojave).
When launching flame I get:
flameTimewarpML bundle payload has already been truncated during previous install. Please copy the original file and start again. Script file location:
/opt/Autodesk/shared/python/flameTimewarpML.py
When I try and use it, I also get this:
python3: can’t open file ‘/Users/paulround/Documents/flameTimewarpML/bundle/command_wrapper.py’:
[Errno 2] No such file or directory
Any clues as to what I’m doing wrong?
I’m no expert but it sounds like you need to re-install it, since it was truncated in the previous attempt
I’ve tried reinstalling several times, I get the same error each time
I’ve seen mentions of having to uninstall it all before re-installing. Have you tried that?
Other than deleting the files, I’m not sure what else I’d need to do.
Sorry, I’m not sure if there’s a more thorough way
It also saves a file in the Mac library - path is in the Readme. Maybe clear all 3 locations and see if it works installing the mac specific one (about a year old) instead?
Do you know which version that would be?
Not sure but if you go down the list from the top it’s the first Mac one you come to I think.
DW, found it,(v0.4.3 2), installed and working!
Hi All. Running into this error:
“cannot pickle ‘WeakMethod’ object in <class ‘TypeError’>”
Any insight would be appreciated! Had to do a bit of manual installing to get this far so may have screwed something up.
If you need someone to test it for you, let me know… …Is it as ‘simple’ as recompiling PyTorch for m1? Or does it involve more?
Thanks!
Hi!
Maybe this has been said, but I´d like report that TWML does not start if the path selected to export the cache has spaces. yesterday this almost drove me crazy before of realize of that issue. I work on linux.
Reviving this post… but wondering if anyone has got this working on the Mac M1s? I would REALLY love to have other timewarp options than Flame’s built in options.
I’m using MLTimewarp v2.4 on mac studio with M1. It works but it is extremely slow compared to a linux box.
The backend TimewarpML uses is available for M1 with native acceleration but it is fairly difficult to distribute using the current way.
I must admit I didn’t have much time recently to move on with TimewarpML though I still hope to be able to find some.Not sure if someone around would like to join-in, but here is several things that has to be done in order to move on:
-
There are still some Ruby code from Julik related to interpolation of he timewarp curves that should to be transferred to python. There’s nothing too difficult there though, just need some time to implement and check.
-
Need to move from OpenCV to native OpenEXR libraries for exr read/write. OpenCV has dropped exr support due to some vulnerabilities.
-
The whole miniconda and pytorch stuff is overkill, it was fast to implement this way but it should be enough to just make single static compiled binary for inference and use this.
-
The rational timewarp between frames is currently achieved by using iterations, and it is highly time-consuming. There is a way to do it in one pass and it should be implemented this way.
I’ve been also doing some tests to do some sort of bespoke training for additional network in order to achive more realistic result for the very shot that’s being timewarped and I think it has some potential as well.
So, long story short, it is possible to get native M1 support for Timewarp when the first three points mentioned above is completed.
I’ll try to find some time for that and if someone wants to help please join in.
Thank you,
Andriy.
I have no idea how to do any of this. But, I want to help.
@talosh, is this the kind of thing that you need a flame artist that also knows these tools, or can we hire someone to help? If we could come up with some funds, would that be helpful? Do you know the right developer that could help?
@daveoy @tombox @panisset @MikeV @ManChicken
Timewarp ML is one of the most important tools to hit flame in its existence. Let’s do this guys.
I don’t have expertise in any of those words talosh said but I would contribute funds.
happy to help from an open-source governance perspective. i can also help with small contributions re: operations / docs / whatever.
as far as core functionality goes: its not my wheelhouse, but i may be able to help migrate opencv apis to openexr, or work on things like that.
if you’re looking for creative input or feedback though, in that respect i’m nothing more than an engineer looking to help out with an open source project.