Support torch.compile in enet, fbnet and 3d-unet pytorch samples #189
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
* support LCM int8 * fix LCM int8-bf16 * fix LCM int8 * update README
… files (#2022) * ipex/efficientnet: extract get_system_config to separate module * ipex/efficientnet: move system_config.py to common folder * common: rename system_config.py to js_sysinfo.py * common: add js_merge.py js_merge is a tool to merge few .json output files together preserving all unique values. Values type mismatch is considered a fatal error. With few values of the same type for the same key only one of the first input is kept with the warning printed out. * common/sysinfo: collect docker info * common/sysinfo: collect dkms info * common/sysinfo: collect dpkg info for key packages * common: add readme * common/sysinfo: add svr-info support * common/sysinfo: differentiate more between docker and baremetal * common/sysinfo: break get_docker_info into 2 functions * common/sysinfo: report key hardware information * common/sysinfo: add lshw support and fetch memory info * common/sysinfo: parse lshw to get cpu info * core/sysinfo: expand cpu info configuration * common/sysinfo: expand configuration for memory info * common/sysinfo: add -o option and improve messaging * common: update readme for sysinfo --------- Signed-off-by: Dmitry Rogozhkin <[email protected]>
* make num_iter flexbile * bugfix for bert-large ddp * bkc for rn50 ddp training update * bkc for rn50 ddp training update * bkc for dlrm_v1 ddp training update * bugfix for llms output --------- Co-authored-by: mahathis <[email protected]>
Signed-off-by: Dmitry Rogozhkin <[email protected]>
* Imported and ran through image recognition and language modelling for Tensorflow CPU workloads Co-authored-by: Mahathi Vatsal <[email protected]>
Signed-off-by: Dmitry Rogozhkin <[email protected]>
* enable yolov5 on CPU Co-authored-by: nick.camarena <[email protected]> Co-authored-by: Clayne Robison <[email protected]> Co-authored-by: Jitendra Patil <[email protected]>
model script runer and setup shell script readme helper to get dataset test for container
Run as:
sudo \
IMAGE=enet \
OUTPUT_DIR=/tmp/output \
PROFILE=$(pwd)/models_v2/pytorch/efficientnet/inference/gpu/profiles/b0.bf16.csv \
PYTHONPATH=$(pwd)/models_v2/common
$(pwd)/models_v2/pytorch/efficientnet/inference/gpu/benchmark.sh
This commit also adds dummy and framework fields to efficientnet
results output and fixes stdev naming in couple places.
Co-authored-by: Voas, Tanner <[email protected]>
Signed-off-by: Dmitry Rogozhkin <[email protected]>
Signed-off-by: Voas, Tanner <[email protected]>
Added new tool json_to_csv to dump multiple json objects to a single CSV (will serialize the json objects) Signed-off-by: Voas, Tanner <[email protected]> Signed-off-by: Dmitry Rogozhkin <[email protected]> Co-authored-by: Voas, Tanner <[email protected]>
Signed-off-by: Dmitry Rogozhkin <[email protected]>
Align summary_utils to efficientnet
Combine functions for dummy / non-dummy inputs
* remove ipex for inductor * fix calibration no prompt for inductor * add fp16 for llm * llm torch.compile forward only
- use descriptive variable names iteration and test rather than i and t in the loop - remove unwanted try/catch statements in common code Signed-off-by: Voas, Tanner <[email protected]> Co-authored-by: mahathis <[email protected]>
* Add dummy mode to swin transformer * Random data, no dataset needed in dummy mode
This patch adds some external metadata to benchmark results. Signed-off-by: Dmitry Rogozhkin <[email protected]>
* update bkc for pvc itex 2.15.0.0 * update bkc for atsm itex 2.15.0.0 * TF 2.15.0 Flex containers (#2087) * validate flex 170 and 140 * Updated baremetal for itex 2.15 (#2098) --------- Co-authored-by: XumingGai <[email protected]> Co-authored-by: Srikanth Ramakrishna <[email protected]> Co-authored-by: Mahathi Vatsal <[email protected]>
- add new telemetry.py tool for capturing telemetry
- Start SMI telemetry capture as its own process inside benchmark.py
- Support UNIX socket communication and python multiprocessing PIPEs
for external control of telemetry start, stop, and termination
- Add requirements to efficientnet sample to work with this
- Add processing code to convert the output CSV into a JSON file
- mentioned metadata in benchmark.py readme
UNIX socket API implemented by Dmitry Rogozhkin <[email protected]>
UNIX socket API adapted into commit by Voas, Tanner <[email protected]>
Co-authored-by: Dmitry Rogozhkin <[email protected]>
Signed-off-by: Voas, Tanner <[email protected]>
Signed-off-by: Voas, Tanner <[email protected]>
- Core inference has been moved to its own class Inference.
- Replaced "NUM_IMAGES" and "NUM_ITERATIONS" with single param "NUM_INPUTS"
- Aligns with other samples usage model (IFRNet, RIFE)
- "NUM_INPUTS" is functionally same as old "NUM_IMAGES" was
- "NUM_ITERATIONS" is 1 in accuracy mode and is dynamic in benchmark
mode based on specified min/max test durations.
- Added support to PyTorch EfficientNet samples to specify min and max test duration
- Logs raw perf with finer granularity now since we have it available
- Use test duration in benchmark for enet
- Remove unused quantization code paths from code
Signed-off-by: Voas, Tanner <[email protected]>
* gpt-j mlperf model
…(#2370) Bumps [torch](https://github.com/pytorch/pytorch) from 1.13.1 to 2.2.0. - [Release notes](https://github.com/pytorch/pytorch/releases) - [Changelog](https://github.com/pytorch/pytorch/blob/main/RELEASE.md) - [Commits](pytorch/pytorch@v1.13.1...v2.2.0) --- updated-dependencies: - dependency-name: torch dependency-type: direct:production ... Signed-off-by: dependabot[bot] <[email protected]>
…ws (#2374) Bumps [github/codeql-action](https://github.com/github/codeql-action) from 3.25.13 to 3.25.15. - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](github/codeql-action@v3.25.13...v3.25.15) --- updated-dependencies: - dependency-name: github/codeql-action dependency-type: direct:production update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
…(#2375) Bumps [ossf/scorecard-action](https://github.com/ossf/scorecard-action) from 2.3.3 to 2.4.0. - [Release notes](https://github.com/ossf/scorecard-action/releases) - [Changelog](https://github.com/ossf/scorecard-action/blob/main/RELEASE.md) - [Commits](ossf/scorecard-action@v2.3.3...v2.4.0) --- updated-dependencies: - dependency-name: ossf/scorecard-action dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <[email protected]>
* refactor gpu * refactor tf * add tf max-gpu * respond to lint errors * remove max-gpu folder * add cuda models to pytorch
Bumps [keras](https://github.com/keras-team/keras) from 2.6.0rc3 to 2.13.1rc0. - [Release notes](https://github.com/keras-team/keras/releases) - [Commits](keras-team/keras@v2.6.0-rc3...v2.13.1-rc0) --- updated-dependencies: - dependency-name: keras dependency-type: direct:production ... Signed-off-by: dependabot[bot] <[email protected]>
…ows (#2382) * Bump super-linter/super-linter from 6.7.0 to 6.8.0 in /.github/workflows Bumps [super-linter/super-linter](https://github.com/super-linter/super-linter) from 6.7.0 to 6.8.0. - [Release notes](https://github.com/super-linter/super-linter/releases) - [Changelog](https://github.com/super-linter/super-linter/blob/main/CHANGELOG.md) - [Commits](super-linter/super-linter@v6.7.0...v6.8.0) --- updated-dependencies: - dependency-name: super-linter/super-linter dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <[email protected]> * Resolved super linter issues --------- Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: Mahathi Vatsal <[email protected]>
* Added new bkm=c 2.4 changes
* specify the version of sympy (#2373)
…dated (#2389) * Removed all instances of miniconda, unvalidated * fixed miniforge base image path
* Removed all instances of miniconda, unvalidated *Also replaced all intel channel to the specific software repo link
Getting 217 fps on fp32 with torch.compile vs. 187 on eager mode running on PVC. Signed-off-by: Dmitry Rogozhkin <[email protected]>
Getting 300 fps on fp32 with torch.compile vs. 187 on eager mode running on PVC. Signed-off-by: Dmitry Rogozhkin <[email protected]>
Signed-off-by: Dmitry Rogozhkin <[email protected]>
dvrogozh
requested review from
ashahba,
claynerobison,
jitendra42,
lerealno and
Mahathi-Vatsal
as code owners
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.
No description provided.