Flame Machine Learning Timewarp, now on Linux and Mac

FYI - running the pip install command also got my A6000 GPU rendering working in 2022.2 on 7.6 as well! (for anyone not on 8.2 yet)

1 Like

Really great using the Beta for ML with Flame Timewarp. Doing a convert from 23.976 to 30FPS. Worked out really well. One caveat is that the EXRs that were returned only had the name on the first EXR. The rest came in with numbers - 002, 003, 004. Not sure if this is expected.

So are you able to set any % you want for the respeed?

Yup. It’s based on flame soft effect time warp settings.

How do i set the amount? When i try its just in 1/2, 1/4 etc

Stick flames timewarp on it at whatever you want, so like 66%. Then when you right click the clip select the ml setting that says flames beta timewarp i think. Sorry not on the box

2 Likes

TWML reads the Timewarp Soft Effect on the clip itself when you choose Timewarp ML / Timewarp from Flame’s TW Effect

2 Likes

This stuff is sorcery. You have saved me so much time this week. I cant thank you enough!

1 Like

Hi. Love that tool. No way to thank Talosh enough for that. A quick question still. After installing it (0.4.3) on a linux system, I have error messages each time à Start Flame. I have to mention that TW works perfectly but I have these 2 types of warnings in the shell when starting:
1st:
Python encountered ModuleNotFoundError (“No module named ‘torch’”) in unknown code
"[PYTHON HOOK] an error occurred. Ignoring python hook…common_quantized.py or common_distributed.py, or hypothesis_utils.py…

2nd:
“[PYTHON HOOK] Duplicate module name found (/opt/Autodesk/shared/python/FlameTimewarpML403/miniconda/lib/python3.8. vs /opt/Autodesk/shared/python/FlameTimewarpML403/miniconda/pkg/python-3.8.-h7579374_1/lib/python3.8”

So basically, it founds 2 versions of the same py hook. Ive got the same lines for a few of them (types.py, typing/py, zip import.py

Should I do something about that (removing some folders or so)?

Trying to get the Konsole app installed, but I get an error “Failed to download metadata for repo ‘AppStrea’: Cannot prepare internalmirrorlist:No URLs in mirrorlist” Is there a fix for this? Running centos 8.2 and 2x A6000.

So that error message came about after Centos 8.2 was end of lifed. I don’t have the solution in front of me, but, some Google Fu should sort you out. You basically need a few commands to update the repositories that yum is using to pull packages from. As in, the repositories that yum pulled from have been marked as end of life and are no longer active.

So, google around with your centos 8.2 and error message and find a solution that seems simple.

Hey @talosh I’ve just installed this and it installed in my home directory. Is there a way to uninstall so I can reinstall somewhere else? Looking forward to trying this on my current job!

Just delete the folder from your home directory, delete any remnants of the Python script in opt autodesk shared python, and reinstall.

1 Like

it sits for a really long time with my terminal windows open but not returning the prompt - its just unzipping something right?

Perhaps a minute or two, but that’s it.

Hi,
I need to run
pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html
but my system is not connected to the internet. What do I need to sort this manually?
ta
D

Hey @talosh

never got around to thanking you for this, it’s fantastic. one question: is there a reason we cannot use a dialog to just specify the percentage to make a non uniform speed change (eg: 115%) instead of the doing it as a timeline fx and then extracting?

thanks for all the hard work here.

Tim

1 Like

Hi folks,
I want to install TimewarpML on a MacBook Pro M1 max but I’ve never installed any Python tools before.
I’m running Flame 2022.3 for work & also have just installed Flame 2023.
Is there a guide to installing TW ML on a mac? I’ve searched this thread but I can only find references to Linux installation

1 Like

It’s the same procedure so you can follow that. There’s a Logik Live for it as well.

1 Like