Skip to content
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

Fix/snippets/dynamism/loop emitters #227

Conversation

a-sidorova
Copy link
Owner

Details:

  • item1
  • ...

Tickets:

  • ticket-id

### Details:
 - *item1*
 - *...*

### Tickets:
 - *ticket-id*

---------

Co-authored-by: Maciej Smyk <[email protected]>
Co-authored-by: Karol Blaszczak <[email protected]>
akladiev and others added 3 commits August 21, 2024 14:41
…otoolkit#25500)

### Details:
- *The PR enables dynamic FP32 MHA tokenization on x64 platforms 🎉*
- *`std::vector.resize()` which was used for buffer scratchpad
allocation is very expensive operation due to default constructor of
elements. This PR replace `std::vector.resize()` with CPU Node
Scratchpad memory which can be shared between nodes. Also since each
thread must have the own scratchpad memory, we allocated `size *
threads_max` - however, in execution thread count can be less (depends
on parallel work amount). Now we allocate only `size * n_threads` where
`nthreads` is real count of working threads.*
- *Fixed dimension K validation in `BrgemmBlocking` pass: one of inputs
can have dynamic value of this dimension*
- *Fixed `utils::broadcast_merge_dim()` and supported broadcasting of
integer values in IterHandlers. Added unit tests for
`utils::broadcast_merge_dim()`*

### Tickets:
 - *149900*


### Prerequisites:
- [x] openvinotoolkit#25326
- [x] openvinotoolkit#25378
- [x] openvinotoolkit#25623
- [x] openvinotoolkit#25638
- [x] openvinotoolkit#25745
- [x] openvinotoolkit#25957
- [x] openvinotoolkit#25733
…lkit#26087)

### Details:
- OneDNN Matmul Brgemm kernel contains wrong debug assertion which lead
to functional failurs on systems with avx2_vnni support.

### Tickets:
 - [CVS-149880](https://jira.devtools.intel.com/browse/CVS-149880)
@a-sidorova a-sidorova force-pushed the fix/snippets/dynamism/loop_emitters branch from fa36a30 to 9c27c6f Compare August 21, 2024 13:16
@github-actions github-actions bot added github_actions Pull requests that update GitHub Actions code category: CI category: docs labels Aug 21, 2024
almilosz and others added 14 commits August 21, 2024 13:34
### Details:
 - Remove testing models from repository
- Add `setup.js` script that will be called once to download testing
models
- Before running the suite check if model is downloaded, add top
describe to *.test.js files

### Notes:
Currently there is no global setup file for native [node test
runner](nodejs/node#49732) as in other test
runner e.g.
[jest](https://jestjs.io/docs/configuration#globalsetup-string)

### Tickets:
 - [CVS-146345](https://jira.devtools.intel.com/browse/CVS-146345)

---------

Co-authored-by: Vishniakov Nikolai <[email protected]>
…lugin (openvinotoolkit#25598)

### Details:
- *Move INFERENCE_PRECISION_HINT test to optional conformance for
meta-plugin*

### Tickets:
 - *147702*

---------

Co-authored-by: Chen Peter <[email protected]>
…oolkit#26150)

### Details:
- GPU tests are only enabled for U20. Checking out the latest oneDNN via
https://github.com/openvinotoolkit/openvino/blob/0b8eb87e0efa9030be47d661fe94b3c8beb655fd/.github/workflows/job_build_linux.yml#L71
breaks the build on U22: [like this
one](https://github.com/openvinotoolkit/openvino/actions/runs/10481070002/job/29030046664)
and prevents other nightly tests from executing.
- This PR limits the checkout of the latest oneDNN to U20 only. This
will unblock the nightly tests that run on U22 for now.
- Later, it might be better to separate the builds for GPU/other test
jobs.

### Tickets:
 - *149805*
…penvinotoolkit#26159)

### Details:
- This issue occurs after
openvinotoolkit#26032
 - Fix segfault in layer tests for onnx_tests.test_lstm.TestLSTM

### Tickets:
 - [CVS-150221](https://jira.devtools.intel.com/browse/CVS-150221)
### Details:
Commit ad4eb09 introduced dynamic loading of ze_loader so that it is
loaded only when really needed.
Recent changes in `ZeroRemoteTensor` caused NPU plugin binary to import
from ze_loader directly.

This commit should fix that and the NPU binary should have no longer
dependency on ze_loader.

### Tickets:
 - None
…6165)

**Details:** Document get_input_by_reference better

**Ticket:** TBD

Signed-off-by: Kazantsev, Roman <[email protected]>
### Details:
 - *Enable fc 4d for MatMul to calculate 4x2*

### Tickets:
 - *132334*
…t#26151)

**Details:** Stabilize L2Loss layer tests on all platforms

**Ticket:** 104863

---------

Signed-off-by: Kazantsev, Roman <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.