Hi Team,
We are trying to run our workflow in GPU,...
# flyte-supporth
helpful-church-28990
09/14/2023, 6:46 AMHi Team,
We are trying to run our workflow in GPU, but when we are running the flyte task we are getting following error:
Copy code
File "/opt/venv/lib/python3.9/site-packages/torch/distributed/distributed_c10d.py", line 1024, in _new_process_group_helper
backend_class = ProcessGroupNCCL(backend_prefix_store, group_rank, group_size, pg_options)
Message:
ProcessGroupNCCL is only supported with GPUs, no GPUs found! Copy code
INFO:dummy.test_gpu:GPU is available in your machine with following device count : 1
2023-09-13 17:00:31,611 [dummy.test_gpu] [INFO ] GPU is available in your machine with following device count : 1
GPU available: True (cuda), used: Trueg4dn.xlargef
freezing-airport-6809
09/14/2023, 1:17 PM
You can specify instance through labels
freezing-airport-6809
09/14/2023, 1:17 PM
You shouldn’t do that inverse pattern
freezing-airport-6809
09/14/2023, 1:18 PM
But coming soon to a Flyte near you us gpu type selection- cc @freezing-boots-56761
freezing-airport-6809
09/14/2023, 1:18 PM
@helpful-church-28990 can you share your snippet of config in the task decorator
h
helpful-church-28990
09/14/2023, 1:24 PM Copy code
@task(requests=Resources(cpu="100m", mem="1Gi", gpu="1"),
limits=Resources(cpu="200m", mem="6Gi", gpu="1"),
)
def test_run(f
freezing-airport-6809
09/14/2023, 1:26 PM
Ohh then why are you using distributed training?
freezing-airport-6809
09/14/2023, 1:26 PM
You don’t need a process group
h
helpful-church-28990
09/14/2023, 1:26 PMSorry what do you mean by distributed training (I am new bie to this field)
f
freezing-airport-6809
09/14/2023, 1:27 PM
If you want to still set it up - my recommendation use flytekit elastic plugin
freezing-airport-6809
09/14/2023, 1:28 PM
Vipul - no worries- single gpu vs multi gpu training is different- from the error it seems your code is trying to setup multi gpu
h
helpful-church-28990
09/14/2023, 1:30 PMOk but in that case i dont think at this moment we need to use multi gpu. So as per my understanding you are suggesting to use to
flytekit elastic pluginf
freezing-airport-6809
09/14/2023, 1:44 PM
It will scale from one to n
freezing-airport-6809
09/14/2023, 1:45 PM
And I agree you shouldn’t need to use more than one
freezing-airport-6809
09/14/2023, 1:45 PM
But then you have to modify your code
h
helpful-church-28990
09/14/2023, 1:51 PMBut then you have to modify your codeThats fine we can do that. but would it be possible to share about how can i define the instance type in flyte recources section
t
tall-lock-23197
09/15/2023, 8:56 AM
You cannot specify GPU type in the task resources. That's still being worked on: https://github.com/flyteorg/flyte/discussions/3796.
If you aren't using distributed training, not sure why your code is hitting that specific code. Are you running an example from the docs?
h
helpful-church-28990
09/15/2023, 9:00 AMI am running the code provided by our Data Scientist, but i suspect is that i am using number of worker as 4 (but as a suspect)
t
tall-lock-23197
09/15/2023, 9:08 AM
Could you share your configuration?
h
helpful-church-28990
09/15/2023, 10:46 AM Copy code
metadata:
name: local
phase: serve
training:
accelerator: gpu
num_workers: 4
hyperparameters:
batch_size: 4096
embedding_dim: 10
weight_decay: 1e-6
lr: 0.005
optim_func: XXX
loss_func: XXX
max_epochs: 10t
tall-lock-23197
09/15/2023, 1:53 PM
I meant, are you using any plugin?
2 Views