chore: ⬆️ Update mudler/parakeet.cpp to 843600590f96a31467a5199f827c253f34c110f7#10198
Merged
Conversation
Update PARAKEET_VERSION to mudler/parakeet.cpp@8436005 (merge of parakeet.cpp#9). Brings NeMo rel_pos_local_attn banded/Longformer attention with the chunk-matmul construction: long audio now uses O(T*window) attention instead of global O(T^2), fixing the encoder OOM on long clips (~16.6-min clip: 54GB->9.4GB peak, ~4x faster) at NeMo's full [128,128] window. Short clips are unchanged (global path). No C-ABI change. Signed-off-by: Ettore Di Giacinto <mudler@localai.io> Assisted-by: Claude:claude-opus-4-8
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Bumps the parakeet-cpp backend pin (
PARAKEET_VERSIONinbackend/go/parakeet-cpp/Makefile) to mudler/parakeet.cpp@843600590, the merge of parakeet.cpp#9.What this brings
NeMo
rel_pos_local_attnbanded / Longformer self-attention with a chunk-matmul construction in the FastConformer encoder. Long audio now uses O(T·window) attention instead of global O(T²), fixing the encoder OOM on long clips that took down a node.Measured on a 16.6-min clip (
tdt-0.6b-v3, GB10, CPU/16t):~4× faster, ~5.7× less peak memory, at NeMo's full
[128,128]window. Banding auto-enables past >8192 encoder frames; short clips stay on the global path and are byte-identical. No C-ABI change (the bump only changes encoder internals).Follow-up
After merge + image rebuild, the parakeet backend can safely handle long audio without OOMing the node.