Fix wrong inputs/model placement when using a single core #725
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
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. |
dacorvo
reviewed
Oct 28, 2024
| move_trace_to_device(model, device_id) | ||
| if is_neuronx_available(): | ||
| torch_neuronx.experimental.set_neuron_cores( | ||
| model, start_nc=0, nc_count=1 |
There was a problem hiding this comment.
What happens if the model has tp_degree > 1 ?
There was a problem hiding this comment.
There is no tp support right now for SD models, only one neuron device is supported, and we want both cores to be used, we apply ddp with the entire model loaded on both cores.
|
This PR is stale because it has been open 15 days with no activity. Remove stale label or comment or this will be closed in 5 days. |
dacorvo
approved these changes
Nov 15, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
It's a recent find that I did not experience before. Previously, when only a single neuron core was used, input tensors were allocated to the same core that the model has been loaded with
torch.jit.load. But it seems that now the input tensors are always allocated to nc:0, whereas the model is loaded to a core with lower usage, so we could bump into unexpected flaky situations where they are in different cores...Before submitting