Zero Collision Hash Benchmark Framework #3127
Open
Conversation
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
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 6, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
729264d to
f95b8fa
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 6, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
f95b8fa to
c44d954
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 6, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
c44d954 to
1a691d5
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
force-pushed
the
export-D77033290
branch
from
1a691d5 to
61b76e8
Compare
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 6, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 6, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 8, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
61b76e8 to
115412b
Compare
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 8, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
force-pushed
the
export-D77033290
branch
from
115412b to
c3663f5
Compare
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 9, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 9, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 14, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 14, 2025
Summary: Pull Request resolved: pytorch#3127 Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
c3663f5 to
e7c1596
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 15, 2025
Summary: Pull Request resolved: https://github.com/pytorch/torchrec/pull/3127M## Context ### Context The high-level intention of this diffs is to open-source (OSS) the Benchmark test bed for hash collision management algorithms. The goal is to provide a testbed for accurate and fair Benchmark. ### Major Changes **Summary** This diff introduces a new benchmark testbed for hash collision management modules. The framework includes several new files and modules: * `torchrec/distributed/benchmark/benchmark_zch/models`: a folder to keep configuration files and wrapper classes for models to benchmark. * `torchrec/distributed/benchmark/benchmark_zch/data`: a folder to keep configuration files and wrapper classes for dataset used for benchmark. * `torchrec/modules/mc_adapter.py`: a new module that implements the MC Adapter algorithm which enables hash collision management modules into embedding collection modules of existing and future models in a plug-and-play manner. The adapter simulates all the APIs of embedding collection and embedding bag collection modules, with a managed collision module being called before embedding look-up. * `torchrec/distributed/benchmark/benchmark_zch/data`: a new module for data-related functions, currently empty. * `torchrec/distributed/benchmark/benchmark_zch/benchmark_zch.py`: a new script that runs the ZCH benchmark. * `torchrec/distributed/benchmark/benchmark_zch/plots`: a new module for plotting training metrics, including an example notebook `plot_training_metrics.ipynb`. The diff includes a significant amount of new code, including model definitions, data loading, and plotting functions. The `benchmark_zch.py` script is the main entry point for running the ZCH benchmark. ### Key Features * Implements the MC Adapter algorithm for ZCH models. * Includes a new benchmark framework for evaluating ZCH models. * Provides data loading and plotting functions for training metrics. ### Implications This diff provides a comprehensive framework for evaluating and optimizing ZCH models. The MC Adapter algorithm is a key component of this framework, and the benchmark script provides a convenient way to run and compare different ZCH models. The plotting functions allow for easy visualization of training metrics, facilitating model optimization and improvement. Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
e7c1596 to
7556c4e
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 15, 2025
Summary: Pull Request resolved: https://github.com/pytorch/torchrec/pull/3127M## Context ### Context The high-level intention of this diffs is to open-source (OSS) the Benchmark test bed for hash collision management algorithms. The goal is to provide a testbed for accurate and fair Benchmark. ### Major Changes **Summary** This diff introduces a new benchmark testbed for hash collision management modules. The framework includes several new files and modules: * `torchrec/distributed/benchmark/benchmark_zch/models`: a folder to keep configuration files and wrapper classes for models to benchmark. * `torchrec/distributed/benchmark/benchmark_zch/data`: a folder to keep configuration files and wrapper classes for dataset used for benchmark. * `torchrec/modules/mc_adapter.py`: a new module that implements the MC Adapter algorithm which enables hash collision management modules into embedding collection modules of existing and future models in a plug-and-play manner. The adapter simulates all the APIs of embedding collection and embedding bag collection modules, with a managed collision module being called before embedding look-up. * `torchrec/distributed/benchmark/benchmark_zch/benchmark_zch.py`: the main entrance of the benchmark testbed. * `torchrec/distributed/benchmark/benchmark_zch/plots`: a folder that keeps plotting notebooks for training and evaluation metrics. ### Key Features * Implements the MC Adapter algorithm for ZCH models. * Includes a new benchmark framework for evaluating hash collision management models. * Provides data loading and plotting functions for training metrics. * Metrics will be output to tensorboard during training for users to inspect the real-time results. ### Implications This diff provides a comprehensive framework for evaluating and optimizing hash collision management models. The MC Adapter algorithm is a key component of this framework, and the benchmark script provides a unified, convenient way to run and compare different hash collision management models. The plotting functions allow for easy visualization of training metrics, facilitating model optimization and improvement. Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
7556c4e to
0970774
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 16, 2025
Summary: Pull Request resolved: https://github.com/pytorch/torchrec/pull/3127M## Context ### Context The high-level intention of this diffs is to open-source (OSS) the Benchmark test bed for hash collision management algorithms. The goal is to provide a testbed for accurate and fair Benchmark. ### Major Changes **Summary** This diff introduces a new benchmark testbed for hash collision management modules. The framework includes several new files and modules: * `torchrec/distributed/benchmark/benchmark_zch/models`: a folder to keep configuration files and wrapper classes for models to benchmark. * `torchrec/distributed/benchmark/benchmark_zch/data`: a folder to keep configuration files and wrapper classes for dataset used for benchmark. * `torchrec/modules/mc_adapter.py`: a new module that implements the MC Adapter algorithm which enables hash collision management modules into embedding collection modules of existing and future models in a plug-and-play manner. The adapter simulates all the APIs of embedding collection and embedding bag collection modules, with a managed collision module being called before embedding look-up. * `torchrec/distributed/benchmark/benchmark_zch/benchmark_zch.py`: the main entrance of the benchmark testbed. * `torchrec/distributed/benchmark/benchmark_zch/plots`: a folder that keeps plotting notebooks for training and evaluation metrics. ### Key Features * Implements the MC Adapter algorithm for ZCH models. * Includes a new benchmark framework for evaluating hash collision management models. * Provides data loading and plotting functions for training metrics. * Metrics will be output to tensorboard during training for users to inspect the real-time results. ### Implications This diff provides a comprehensive framework for evaluating and optimizing hash collision management models. The MC Adapter algorithm is a key component of this framework, and the benchmark script provides a unified, convenient way to run and compare different hash collision management models. The plotting functions allow for easy visualization of training metrics, facilitating model optimization and improvement. Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
0970774 to
6d2e16f
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 17, 2025
Summary: Pull Request resolved: https://github.com/pytorch/torchrec/pull/3127M## Context ### Context The high-level intention of this diffs is to open-source (OSS) the Benchmark test bed for hash collision management algorithms. The goal is to provide a testbed for accurate and fair Benchmark. ### Major Changes **Summary** This diff introduces a new benchmark testbed for hash collision management modules. The framework includes several new files and modules: * `torchrec/distributed/benchmark/benchmark_zch/models`: a folder to keep configuration files and wrapper classes for models to benchmark. * `torchrec/distributed/benchmark/benchmark_zch/data`: a folder to keep configuration files and wrapper classes for dataset used for benchmark. It also includes a pre-hash script `sparse_kuairand_dataset.py` which takes an input of kuairand dataset and make the input values evenly distributed among the input hash space. * `torchrec/modules/mc_adapter.py`: a new module that implements the MC Adapter algorithm which enables hash collision management modules into embedding collection modules of existing and future models in a plug-and-play manner. The adapter simulates all the APIs of embedding collection and embedding bag collection modules, with a managed collision module being called before embedding look-up. * `torchrec/distributed/benchmark/benchmark_zch/benchmark_zch.py`: the main entrance of the benchmark testbed. * `torchrec/distributed/benchmark/benchmark_zch/plots`: a folder that keeps plotting notebooks for training and evaluation metrics. ### Key Features * Implements the MC Adapter algorithm for ZCH models. * Includes a new benchmark framework for evaluating hash collision management models. * Provides data loading and plotting functions for training metrics. * Metrics will be output to tensorboard during training for users to inspect the real-time results. ### Implications This diff provides a comprehensive framework for evaluating and optimizing hash collision management models. The MC Adapter algorithm is a key component of this framework, and the benchmark script provides a unified, convenient way to run and compare different hash collision management models. The plotting functions allow for easy visualization of training metrics, facilitating model optimization and improvement. Reviewed By: aporialiao Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
6d2e16f to
a84406c
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jul 17, 2025
Summary: Pull Request resolved: https://github.com/pytorch/torchrec/pull/3127M## Context ### Context The high-level intention of this diffs is to open-source (OSS) the Benchmark test bed for hash collision management algorithms. The goal is to provide a testbed for accurate and fair Benchmark. ### Major Changes **Summary** This diff introduces a new benchmark testbed for hash collision management modules. The framework includes several new files and modules: * `torchrec/distributed/benchmark/benchmark_zch/models`: a folder to keep configuration files and wrapper classes for models to benchmark. * `torchrec/distributed/benchmark/benchmark_zch/data`: a folder to keep configuration files and wrapper classes for dataset used for benchmark. It also includes a pre-hash script `sparse_kuairand_dataset.py` which takes an input of kuairand dataset and make the input values evenly distributed among the input hash space. * `torchrec/modules/mc_adapter.py`: a new module that implements the MC Adapter algorithm which enables hash collision management modules into embedding collection modules of existing and future models in a plug-and-play manner. The adapter simulates all the APIs of embedding collection and embedding bag collection modules, with a managed collision module being called before embedding look-up. * `torchrec/distributed/benchmark/benchmark_zch/benchmark_zch.py`: the main entrance of the benchmark testbed. * `torchrec/distributed/benchmark/benchmark_zch/plots`: a folder that keeps plotting notebooks for training and evaluation metrics. ### Key Features * Implements the MC Adapter algorithm for ZCH models. * Includes a new benchmark framework for evaluating hash collision management models. * Provides data loading and plotting functions for training metrics. * Metrics will be output to tensorboard during training for users to inspect the real-time results. ### Implications This diff provides a comprehensive framework for evaluating and optimizing hash collision management models. The MC Adapter algorithm is a key component of this framework, and the benchmark script provides a unified, convenient way to run and compare different hash collision management models. The plotting functions allow for easy visualization of training metrics, facilitating model optimization and improvement. Reviewed By: aporialiao Differential Revision: D77033290
lizhouyu
force-pushed
the
export-D77033290
branch
from
a84406c to
9230233
Compare
Summary: Pull Request resolved: https://github.com/pytorch/torchrec/pull/3127M## Context ### Context The high-level intention of this diffs is to open-source (OSS) the Benchmark test bed for hash collision management algorithms. The goal is to provide a testbed for accurate and fair Benchmark. ### Major Changes **Summary** This diff introduces a new benchmark testbed for hash collision management modules. The framework includes several new files and modules: * `torchrec/distributed/benchmark/benchmark_zch/models`: a folder to keep configuration files and wrapper classes for models to benchmark. * `torchrec/distributed/benchmark/benchmark_zch/data`: a folder to keep configuration files and wrapper classes for dataset used for benchmark. It also includes a pre-hash script `sparse_kuairand_dataset.py` which takes an input of kuairand dataset and make the input values evenly distributed among the input hash space. * `torchrec/modules/mc_adapter.py`: a new module that implements the MC Adapter algorithm which enables hash collision management modules into embedding collection modules of existing and future models in a plug-and-play manner. The adapter simulates all the APIs of embedding collection and embedding bag collection modules, with a managed collision module being called before embedding look-up. * `torchrec/distributed/benchmark/benchmark_zch/benchmark_zch.py`: the main entrance of the benchmark testbed. * `torchrec/distributed/benchmark/benchmark_zch/plots`: a folder that keeps plotting notebooks for training and evaluation metrics. ### Key Features * Implements the MC Adapter algorithm for ZCH models. * Includes a new benchmark framework for evaluating hash collision management models. * Provides data loading and plotting functions for training metrics. * Metrics will be output to tensorboard during training for users to inspect the real-time results. ### Implications This diff provides a comprehensive framework for evaluating and optimizing hash collision management models. The MC Adapter algorithm is a key component of this framework, and the benchmark script provides a unified, convenient way to run and compare different hash collision management models. The plotting functions allow for easy visualization of training metrics, facilitating model optimization and improvement. Reviewed By: aporialiao Differential Revision: D77033290
|
This pull request was exported from Phabricator. Differential Revision: D77033290 |
lizhouyu
force-pushed
the
export-D77033290
branch
from
9230233 to
6e3df72
Compare
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 17, 2025
Summary: Pull Request resolved: https://github.com/pytorch/torchrec/pull/3127M## Context ### Context The high-level intention of this diffs is to open-source (OSS) the Benchmark test bed for hash collision management algorithms. The goal is to provide a testbed for accurate and fair Benchmark. ### Major Changes **Summary** This diff introduces a new benchmark testbed for hash collision management modules. The framework includes several new files and modules: * `torchrec/distributed/benchmark/benchmark_zch/models`: a folder to keep configuration files and wrapper classes for models to benchmark. * `torchrec/distributed/benchmark/benchmark_zch/data`: a folder to keep configuration files and wrapper classes for dataset used for benchmark. It also includes a pre-hash script `sparse_kuairand_dataset.py` which takes an input of kuairand dataset and make the input values evenly distributed among the input hash space. * `torchrec/modules/mc_adapter.py`: a new module that implements the MC Adapter algorithm which enables hash collision management modules into embedding collection modules of existing and future models in a plug-and-play manner. The adapter simulates all the APIs of embedding collection and embedding bag collection modules, with a managed collision module being called before embedding look-up. * `torchrec/distributed/benchmark/benchmark_zch/benchmark_zch.py`: the main entrance of the benchmark testbed. * `torchrec/distributed/benchmark/benchmark_zch/plots`: a folder that keeps plotting notebooks for training and evaluation metrics. ### Key Features * Implements the MC Adapter algorithm for ZCH models. * Includes a new benchmark framework for evaluating hash collision management models. * Provides data loading and plotting functions for training metrics. * Metrics will be output to tensorboard during training for users to inspect the real-time results. ### Implications This diff provides a comprehensive framework for evaluating and optimizing hash collision management models. The MC Adapter algorithm is a key component of this framework, and the benchmark script provides a unified, convenient way to run and compare different hash collision management models. The plotting functions allow for easy visualization of training metrics, facilitating model optimization and improvement. Reviewed By: aporialiao Differential Revision: D77033290 fbshipit-source-id: 298a1c6e1ab858641992db7362ccf227725ec12b
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.
Differential Revision: D77033290