-
Notifications
You must be signed in to change notification settings - Fork 2.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[core] Add specific tensor descriptor for Result #25954
Merged
mlukasze
merged 47 commits into
openvinotoolkit:master
from
praasz:feature/specific-tensor-descriptor-for-results
Dec 12, 2024
Merged
[core] Add specific tensor descriptor for Result #25954
mlukasze
merged 47 commits into
openvinotoolkit:master
from
praasz:feature/specific-tensor-descriptor-for-results
Dec 12, 2024
Conversation
This file contains 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
github-actions
bot
added
category: inference
OpenVINO Runtime library - Inference
category: Core
OpenVINO Core (aka ngraph)
category: transformations
OpenVINO Runtime library - Transformations
category: CPP API
OpenVINO CPP API bindings
labels
Aug 7, 2024
github-actions
bot
added
category: TF FE
OpenVINO TensorFlow FrontEnd
category: PyTorch FE
OpenVINO PyTorch Frontend
labels
Aug 19, 2024
…nsor-descriptor-for-results
…nsor-descriptor-for-results
- add UnrollIf to check tensor names are not lost if body model is Parameter -> Result Signed-off-by: Raasz, Pawel <[email protected]>
This was referenced Nov 15, 2024
github-merge-queue bot
pushed a commit
that referenced
this pull request
Nov 20, 2024
### Details: - Enable simple `If` test for CPU plugin, Test pass as tensor names are not checked. - PR #25954 fixes issue with tensor names for `unroll_if` transformation when `If` operator has got simple body like `Parameter -> Result` ### Tickets: - CVS-114763 Signed-off-by: Raasz, Pawel <[email protected]>
alexsu52
approved these changes
Nov 25, 2024
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I confirm that this PR fixes NNCF layer-wise algorithms such as GPTQ.
NishantPrabhuFujitsu
pushed a commit
to NishantPrabhuFujitsu/openvino
that referenced
this pull request
Nov 26, 2024
### Details: - Enable simple `If` test for CPU plugin, Test pass as tensor names are not checked. - PR openvinotoolkit#25954 fixes issue with tensor names for `unroll_if` transformation when `If` operator has got simple body like `Parameter -> Result` ### Tickets: - CVS-114763 Signed-off-by: Raasz, Pawel <[email protected]>
github-merge-queue bot
pushed a commit
that referenced
this pull request
Nov 27, 2024
…pdate shapes, types (#27594) ### Details: - Remove friend of `ReverseShapeAndTypeInfer` from Tensor descriptor class - Setting the shapes in tensor descriptor directly not using the Tensor API can create risk e.g. PartialShape and Shape of Tensor descriptor will not be sync correctly. ### Related PR: - #25954 ### Tickets: - CVS-146649 --------- Signed-off-by: Raasz, Pawel <[email protected]> Signed-off-by: Pawel Raasz <[email protected]>
…nsor-descriptor-for-results
…nsor-descriptor-for-results
jane-intel
approved these changes
Dec 12, 2024
auto-merge was automatically disabled
December 12, 2024 15:37
Pull Request is not mergeable
Merged
via the queue into
openvinotoolkit:master
with commit Dec 12, 2024
8d6491b
174 of 176 checks passed
github-merge-queue bot
pushed a commit
that referenced
this pull request
Dec 13, 2024
### Details: - For models with `version > 10` the node's friendly name and tensor's names will be not moved from original node. If conversion node added by PPP or convert precision then new friendly name will be created based on previous node. The new node name will have format `[previous_node_name].[port_number]`. - Align `ConvertPrecision` transformation to with PPP - The issue with lost tensor names when set them on model's inputss/outputs has been solved in #25954 - For model version 10 the old behavior is preserved as legacy compatibility mode ### Tickets: - [CVS-127482](https://jira.devtools.intel.com/browse/CVS-127482) ### Depends on: - #25954 - implements tensor names handling for model outputs (should be used here) --------- Signed-off-by: Raasz, Pawel <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]> Co-authored-by: Anastasia Kuporosova <[email protected]>
11happy
pushed a commit
to 11happy/openvino
that referenced
this pull request
Dec 23, 2024
### Details: - Introduce shared tensor descriptor: - It has specific tensor names. - When specific tensor names set they will be used as tensor names otherwise the shred tensor names are used. - Specific tensor names will be added to shared descriptor (input descriptor), and transferred to new tensor descriptor if changed. - Use shared tensor descriptor for Result, which got specific tensor names. - Add dev API Extension interface for descriptor::Tensor for internal OV use. - Update `ReverseShapeAndTypeInfer` to use descriptor interface instead change properties directly. Update `PartialShape` of descriptor can change it in undefined state with `Shape` property. - Improves tensor names handling for model outputs introduced in openvinotoolkit#23713 - Fix tensor names issue in `unroll_if` transformation (ticket: 114763) ### Tickets: - CVS-143105 - CVS-146649 - CVS-114763 ### Blocked by: - intel-innersource/frameworks.ai.openvino.tests#1684 - fixes e2e tests --------- Signed-off-by: Raasz, Pawel <[email protected]>
11happy
pushed a commit
to 11happy/openvino
that referenced
this pull request
Dec 23, 2024
…t#23713) ### Details: - For models with `version > 10` the node's friendly name and tensor's names will be not moved from original node. If conversion node added by PPP or convert precision then new friendly name will be created based on previous node. The new node name will have format `[previous_node_name].[port_number]`. - Align `ConvertPrecision` transformation to with PPP - The issue with lost tensor names when set them on model's inputss/outputs has been solved in openvinotoolkit#25954 - For model version 10 the old behavior is preserved as legacy compatibility mode ### Tickets: - [CVS-127482](https://jira.devtools.intel.com/browse/CVS-127482) ### Depends on: - openvinotoolkit#25954 - implements tensor names handling for model outputs (should be used here) --------- Signed-off-by: Raasz, Pawel <[email protected]> Co-authored-by: Michal Lukaszewski <[email protected]> Co-authored-by: Anastasia Kuporosova <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Labels
category: Core
OpenVINO Core (aka ngraph)
category: CPP API
OpenVINO CPP API bindings
category: inference
OpenVINO Runtime library - Inference
category: JAX FE
OpenVINO JAX FrontEnd
category: PyTorch FE
OpenVINO PyTorch Frontend
category: TF FE
OpenVINO TensorFlow FrontEnd
category: transformations
OpenVINO Runtime library - Transformations
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.
Details:
Introduce shared tensor descriptor:
Use shared tensor descriptor for Result, which got specific tensor names.
Add dev API Extension interface for descriptor::Tensor for internal OV use.
Update
ReverseShapeAndTypeInfer
to use descriptor interface instead change properties directly. UpdatePartialShape
of descriptor can change it in undefined state withShape
property.Improves tensor names handling for model outputs introduced in [core] Preserve friendly name and tensor names in PPP #23713
Fix tensor names issue in
unroll_if
transformation (ticket: 114763)Tickets:
Blocked by: