Adding command line option to move model sampling to cpu #6502
+10
−0
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Outdated suggestions cannot be applied.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.
The work being done on the GPU is minimal but requires a full round trip to the GPU and back for each iteration being done. Regular GPUs dont show this as much because of the fast PCIe interconnect. Juice (juicelabs.co) on the other hand, uses ethernet as it's bus and this back and forth turned 139ms to 7s. By moving the model sampling to the CPU, the 7s disappeared.
Hidden behind a command line option to preserve current functionality.