BN-83 GPU support integration #118
Conversation
Signed-off-by: Sylvester La-Tunje <[email protected]> (cherry picked from commit a69c56c)
slatunje
changed the title
There was a problem hiding this comment.
Ok, looks like a simple change but for sure will require generalization and additional logic for AWS & local providers... Will check how to do that.
Also could you please attach some example blend file to test CUDA properly? Something that will clearly show that the CUDA is working, not just CPU
| @@ -491,6 +529,8 @@ def createInstanceAgent(cfg): | |||
| 'description': 'BlendNet Agent worker', | |||
| 'scheduling': { | |||
| 'preemptible': cfg['use_cheap_instance'], | |||
| 'onHostMaintenance': 'TERMINATE', | |||
| 'automaticRestart': True, | |||
There was a problem hiding this comment.
Actually Manager should decide what to do with the agent instance.
There was a problem hiding this comment.
These changes were simply added to allow me to get gpu working.
Feel free to move it to where is fits best as I do not know the code base very well
Thanks. In addition to these logic per provider, there needs to be logic per GPU availability per zone.
I used private blend files from a private project and I also used the barbershop https://www.blender.org/download/demo-files/#cycles for gpu. Barbershop took too long as I only have 1 gpu quota so was unable to test huge files all the way. Only long enough to see the gpu's working on metrics and within |
|
Ok, glad it's working for you - but seems it will need adjustments to fit everywhere, so I think will move these activity to v0.5 |
Signed-off-by: Sylvester La-Tunje <[email protected]>
note this will increase build time Signed-off-by: Sylvester La-Tunje <[email protected]>
Signed-off-by: Sylvester La-Tunje [email protected]
(cherry picked from commit a69c56c)