Repository navigation
Slice gated MoE activations directly and reuse sorted expert IDs. - #813
Open
copybara-service[bot] wants to merge 1 commit into
Open
copybara-service[bot] wants to merge 1 commit into
copybara-service[bot] wants to merge 1 commit into
Conversation
In Gemma4 MoE training profiles, `MoERagged` reshapes `gate_out` from `[B_total, 2 * hidden_dim]` to `[B_total, 2, hidden_dim]` to extract `gate_out[:, 0, :]` and `gate_out[:, 1, :]` (triggering sublane layout copies on the 2-element axis) and recomputes sorted expert indices via `jnp.repeat`. Slice `gate_out[:, :self.hidden_dim]` and `gate_out[:, self.hidden_dim:]` contiguously along the minor dimension in `MoERagged`, and use `jax.lax.sort_key_val` in `_expert_dispatch` to return `sorted_expert_choices` directly instead of recomputing `jnp.repeat`. PiperOrigin-RevId: 996117699
|
Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA). View this failed invocation of the CLA check for more information. For the most up to date status, view the checks section at the bottom of the pull request. |
This file contains hidden or 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Slice gated MoE activations directly and reuse sorted expert IDs.
In Gemma4 MoE training profiles,
MoERaggedreshapesgate_outfrom[B_total, 2 * hidden_dim]to[B_total, 2, hidden_dim]to extractgate_out[:, 0, :]andgate_out[:, 1, :](triggering sublane layout copies on the 2-element axis) and recomputes sorted expert indices viajnp.repeat.Slice
gate_out[:, :self.hidden_dim]andgate_out[:, self.hidden_dim:]contiguously along the minor dimension inMoERagged, and usejax.lax.sort_key_valin_expert_dispatchto returnsorted_expert_choicesdirectly instead of recomputingjnp.repeat.