Changing Backend Tensor initialization #5055
Open
+20
−5
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Outdated suggestions cannot be applied.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.
Summary:
X-link: https://github.com/facebookresearch/FBGEMM/pull/2065
Context:
Currently, RocksDB stores data on row-wise format, to enable optimizer offloading for the Kernel. We will append the optimizer state to its corresponding row.
During initialization, we need to randomly initialize weights while the optimizer values need to initialized to zero.
When optimizer offloading is enabled,
In this diff:
We add two new arguments:
Scenarios:
Optimizer_offloading is False:
max_D = Dimensions of weights only,
optimizer_D = 0
Optimizer_offloading is True:
max_D = Dimension of weights (w_D) + optimizers (o_D)
optimizer_D = dimensions of optimizers (o_D)
initialize o_D columns with zero
Differential Revision: D85157732