Closed michaelbenayoun closed 2 months ago
This PR tries to fix the issue reported here https://github.com/aws-neuron/neuronx-distributed/issues/24.
It seems to be linked to torch.autocast:
torch.autocast
with torch.autocast(enabled=False):
with contexlib.nullcontex():
For some reason the first approach was leading to different cores trying to execute different graphs, leading to the MPMD detected error.
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.
What does this PR do?
This PR tries to fix the issue reported here https://github.com/aws-neuron/neuronx-distributed/issues/24.
It seems to be linked to
torch.autocast
:with torch.autocast(enabled=False):
with contexlib.nullcontex():
.For some reason the first approach was leading to different cores trying to execute different graphs, leading to the MPMD detected error.