[stale] Clean feature history#67
Open
stevenvar wants to merge 16 commits intojoeyye-work:new-serving-2.20from
Open
[stale] Clean feature history#67stevenvar wants to merge 16 commits intojoeyye-work:new-serving-2.20from
stevenvar wants to merge 16 commits intojoeyye-work:new-serving-2.20from
Conversation
Adding --tf_xla_annotate_cluster_id to allow operator name starting with .cluster.id to influcence clustering decisions
Adds the CPU runtime plumbing needed to carry the dynamic batch value through execution and lowers the supporting runtime pieces. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Adds batch matching support and updates output handling so compiled executions can normalize and restore batch-aware values. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Introduces symbolic dimension expressions to TensorFlow and XLA shape representations so dynamic dimensions can be tracked structurally instead of only by concrete values. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Captures dynamic dimension values at compile time and adds clustering safeguards so incompatible dynamic-shape regions are not merged accidentally. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Keeps inferred shape metadata alive when values cross JIT and encapsulation boundaries so downstream compilation can still recover symbolic shape information. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Tightens clustering and shape-inference behavior around partially known or unranked shapes to avoid losing dynamic-shape information in unsupported cases. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Improves tf2xla shape propagation and normalizes padded constants so shape-derived values can continue through lowering in a consistent way. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Fixes the reshape ratio handling so shape-expression information stays consistent when reshape dimensions are derived from symbolic values. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Limits the Eigen fallback path to dynamic CPU dot cases so static cases continue to use the normal lowering while dynamic dimensions remain supported. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Shares common expression-inference helpers across passes so symbolic dimension analysis stays aligned in different parts of the pipeline. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
…operators Extends expression updates across additional TensorFlow operators so symbolic dimensions continue to reflect the same shape transformations as concrete dimensions. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
…mic_sizes Removes the separate single-dynamic-dimension flag and reuses the existing dynamic-sizes flag so the feature is configured in one place. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Cleans up remaining symbolic-expression propagation details and aligns edge cases discovered during integration of the earlier feature commits. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
Fixes shape-expression bookkeeping after removing dimensions so the expression vector stays aligned with the resulting shape. Co-authored-by: Steven Varoumas <steven.varoumas1@huawei.com> Co-authored-by: Utku Saglam <utku.saglam@h-partners.com> Co-authored-by: Jinyun (Joey) Ye <jinyunye@huawei.com> Co-authored-by: Muteages <67578152+Muteages@users.noreply.github.com> Co-authored-by: Guillermo Callaghan <guillermo2@huawei.com>
stevenvar
force-pushed
the
steven.joey_for_serving_history_by_feature_clean
branch
3 times, most recently
from
4b15ce8 to
3eb8a9e
Compare
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
4 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.
No description provided.