Benchamrk framework for torchrec #3072
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: D76150895 |
Summary: Pull Request resolved: pytorch#3017 ### Major changes - Copy the following files from `fb` to corresponding location in the `torchrec` repository - `fb/distributed/hash_mc_embedding.py → torchrec/distributed/hash_mc_embedding.py` - `fb/modules/hash_mc_evictions.py → torchrec/modules/hash_mc_evictions.py` - `fb/modules/hash_mc_metrics.py → torchrec/modules/hash_mc_metrics.py` - `fb/modules/hash_mc_modules.py → torchrec/modules/hash_mc_modules.py` - `fb/modules/tests/test_hash_mc_evictions.py → torchrec/modules/tests/test_hash_mc_evictions.py` - `fb/modules/tests/test_hash_mc_modules.py → torchrec/modules/tests/test_hash_mc_modules.py` - Create a `test_hash_zch_mc.py` file in `torchrec/distributed/tests` folder following the `test_quant_mc_embedding.py` in `torchrec/fb/distributed/tests`. - trimmed quantization and inference codes, and only kept the training part. - rewire the related packages from `torchrec.fb` to `torchrec` - Update `BUCK` files in related folders - Update the affected repos to use `torchrec` modules instead of the modules in `torchrec.fb` - Update `/modules/hash_mc_metrics.py` - Replace the tensorboard module with a local file logger in `hash_mc_metrics.py` module - Update the license declaration headers for the four OSS files ### ToDos after landing this Diff - Clean the duplicated `hash_mc_modules.py` file in the `fb` folder for safe landing. Differential Revision: D76476676
Summary: ### Major changes - Create a `mpzch` folder under the `torchrec/github/examples` folder - Implement a simple SparseArch module with a flag to switch between original and MPZCH managed collision modules - Profile the running time and QPS for model training(GPU)/inference(CPU) - Create a notebook tutorial for ZCH basics and the use of ZCH modules in TorchRec ### ToDos for OSS - When the internal torchrec MPZCH module is OSS - Remove the `BUCK` file - Replace all the `from torchrec.fb.modules` in `sparse_arch.py` to `from torchrec.modules` ### Potential improvement - Add hash collision counter - Show profiling results in the Readme file - Add multi-batch profiling Differential Revision: D75570684 Reviewed By: aporialiao
|
This pull request was exported from Phabricator. Differential Revision: D76150895 |
lizhouyu
force-pushed
the
export-D76150895
branch
from
f06a2f1 to
53a0136
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76150895 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jun 16, 2025
Summary: Pull Request resolved: pytorch#3072 # Benchmark framework for MPZCH ### Major changes - Add a `benchmark prober` in `torchrec/distributed/benchmark/benchmark_zch_utils.py` to collect and calculate the zero collision hash related metrics like hit count, insert count, and collision count. - Implement a `benchmark_zch_dlrmv2` local testbed in `torchrec/distributed/benchmark/benchmark_zch_dlrmv2.py`, which allows to profile a DLRMv2 model with and without the MPZCH enabled, and record the metrics including ZCH-related metrics, QPS, NE, and AUROC. - Add `mc_adapter` modules in `torchrec/modules/mc_adapter.py`. These modules enable seamless replacement of embedding collection and embedding bag collection modules with the managed collision version. - Add two dictionaries `self.table_name_on_device_remapped_ids_dict` and `self.table_name_on_device_input_ids_dict` in the `HashZchManagedCollisionModule` module in `torchrec/modules/hash_mc_modules.py` to record the remapped identities and input feature values to the MPZCH module on current rank respectively after input mapping. - Add `count_non_zch_collision.py` script to count the collision rate of non-zch modules after performing `murmur_hash3`. - Add the criteo kaggle dataset data loader in `torchrec/distributed/benchmark/data` and revise the `hashes` attribute of data pipeline in the `_get_in_memory_dataloader` function in the `torchrec/distributed/benchmark/data/dlrm_dataloader.py` file to pre-hash the input feature values to the passed-in argument `input_hash_size` (defaultly as 100000). - Note that we can change the `single_ttl` in the `HashZchSingleTtlScorer` module of `torchrec/modules/hash_mc_evictions.py` to change the eviticability of identities in each `HashZchManagedCollisionModule` module, since exiting benchmark workflow only takes several minutes on the subset of the criteo-kaggle dataset. By default the identities become evictable after one hour. This descrepency leads to non-eviction during the profiling process. ### Dataset - [Criteo Kaggle Small](https://drive.google.com/file/d/1__rPcUSa45FHkmnBwivuM7K4nMYWD7b7/view?usp=sharing) - [Criteo Kaggle](https://drive.google.com/file/d/1_lAbXTEOk5vlPGXd4UvTrxGV6sCPer_R/view?usp=drive_link) Differential Revision: D76150895
lizhouyu
force-pushed
the
export-D76150895
branch
from
53a0136 to
f815ed2
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76150895 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jun 16, 2025
Summary: Pull Request resolved: pytorch#3072 # Benchmark framework for MPZCH ### Major changes - Add a `benchmark prober` in `torchrec/distributed/benchmark/benchmark_zch_utils.py` to collect and calculate the zero collision hash related metrics like hit count, insert count, and collision count. - Implement a `benchmark_zch_dlrmv2` local testbed in `torchrec/distributed/benchmark/benchmark_zch_dlrmv2.py`, which allows to profile a DLRMv2 model with and without the MPZCH enabled, and record the metrics including ZCH-related metrics, QPS, NE, and AUROC. - Add `mc_adapter` modules in `torchrec/modules/mc_adapter.py`. These modules enable seamless replacement of embedding collection and embedding bag collection modules with the managed collision version. - Add two dictionaries `self.table_name_on_device_remapped_ids_dict` and `self.table_name_on_device_input_ids_dict` in the `HashZchManagedCollisionModule` module in `torchrec/modules/hash_mc_modules.py` to record the remapped identities and input feature values to the MPZCH module on current rank respectively after input mapping. - Add `count_non_zch_collision.py` script to count the collision rate of non-zch modules after performing `murmur_hash3`. - Add the criteo kaggle dataset data loader in `torchrec/distributed/benchmark/data` and revise the `hashes` attribute of data pipeline in the `_get_in_memory_dataloader` function in the `torchrec/distributed/benchmark/data/dlrm_dataloader.py` file to pre-hash the input feature values to the passed-in argument `input_hash_size` (defaultly as 100000). - Note that we can change the `single_ttl` in the `HashZchSingleTtlScorer` module of `torchrec/modules/hash_mc_evictions.py` to change the eviticability of identities in each `HashZchManagedCollisionModule` module, since exiting benchmark workflow only takes several minutes on the subset of the criteo-kaggle dataset. By default the identities become evictable after one hour. This descrepency leads to non-eviction during the profiling process. ### Dataset - [Criteo Kaggle Small](https://drive.google.com/file/d/1__rPcUSa45FHkmnBwivuM7K4nMYWD7b7/view?usp=sharing) - [Criteo Kaggle](https://drive.google.com/file/d/1_lAbXTEOk5vlPGXd4UvTrxGV6sCPer_R/view?usp=drive_link) Differential Revision: D76150895
lizhouyu
force-pushed
the
export-D76150895
branch
from
f815ed2 to
bfd8143
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76150895 |
lizhouyu
added a commit
to lizhouyu/torchrec
that referenced
this pull request
Jun 16, 2025
Summary: Pull Request resolved: pytorch#3072 # Benchmark framework for MPZCH ### Major changes - Add a `benchmark prober` in `torchrec/distributed/benchmark/benchmark_zch_utils.py` to collect and calculate the zero collision hash related metrics like hit count, insert count, and collision count. - Implement a `benchmark_zch_dlrmv2` local testbed in `torchrec/distributed/benchmark/benchmark_zch_dlrmv2.py`, which allows to profile a DLRMv2 model with and without the MPZCH enabled, and record the metrics including ZCH-related metrics, QPS, NE, and AUROC. - Add `mc_adapter` modules in `torchrec/modules/mc_adapter.py`. These modules enable seamless replacement of embedding collection and embedding bag collection modules with the managed collision version. - Add two dictionaries `self.table_name_on_device_remapped_ids_dict` and `self.table_name_on_device_input_ids_dict` in the `HashZchManagedCollisionModule` module in `torchrec/modules/hash_mc_modules.py` to record the remapped identities and input feature values to the MPZCH module on current rank respectively after input mapping. - Add `count_non_zch_collision.py` script to count the collision rate of non-zch modules after performing `murmur_hash3`. - Add the criteo kaggle dataset data loader in `torchrec/distributed/benchmark/data` and revise the `hashes` attribute of data pipeline in the `_get_in_memory_dataloader` function in the `torchrec/distributed/benchmark/data/dlrm_dataloader.py` file to pre-hash the input feature values to the passed-in argument `input_hash_size` (defaultly as 100000). - Note that we can change the `single_ttl` in the `HashZchSingleTtlScorer` module of `torchrec/modules/hash_mc_evictions.py` to change the eviticability of identities in each `HashZchManagedCollisionModule` module, since exiting benchmark workflow only takes several minutes on the subset of the criteo-kaggle dataset. By default the identities become evictable after one hour. This descrepency leads to non-eviction during the profiling process. ### Dataset - [Criteo Kaggle Small](https://drive.google.com/file/d/1__rPcUSa45FHkmnBwivuM7K4nMYWD7b7/view?usp=sharing) - [Criteo Kaggle](https://drive.google.com/file/d/1_lAbXTEOk5vlPGXd4UvTrxGV6sCPer_R/view?usp=drive_link) Differential Revision: D76150895
lizhouyu
force-pushed
the
export-D76150895
branch
from
bfd8143 to
8036d23
Compare
Summary: Pull Request resolved: pytorch#3072 # Benchmark framework for MPZCH ### Major changes - Add a `benchmark prober` in `torchrec/distributed/benchmark/benchmark_zch_utils.py` to collect and calculate the zero collision hash related metrics like hit count, insert count, and collision count. - Implement a `benchmark_zch_dlrmv2` local testbed in `torchrec/distributed/benchmark/benchmark_zch_dlrmv2.py`, which allows to profile a DLRMv2 model with and without the MPZCH enabled, and record the metrics including ZCH-related metrics, QPS, NE, and AUROC. - Add `mc_adapter` modules in `torchrec/modules/mc_adapter.py`. These modules enable seamless replacement of embedding collection and embedding bag collection modules with the managed collision version. - Add two dictionaries `self.table_name_on_device_remapped_ids_dict` and `self.table_name_on_device_input_ids_dict` in the `HashZchManagedCollisionModule` module in `torchrec/modules/hash_mc_modules.py` to record the remapped identities and input feature values to the MPZCH module on current rank respectively after input mapping. - Add `count_non_zch_collision.py` script to count the collision rate of non-zch modules after performing `murmur_hash3`. - Add the criteo kaggle dataset data loader in `torchrec/distributed/benchmark/data` and revise the `hashes` attribute of data pipeline in the `_get_in_memory_dataloader` function in the `torchrec/distributed/benchmark/data/dlrm_dataloader.py` file to pre-hash the input feature values to the passed-in argument `input_hash_size` (defaultly as 100000). - Note that we can change the `single_ttl` in the `HashZchSingleTtlScorer` module of `torchrec/modules/hash_mc_evictions.py` to change the eviticability of identities in each `HashZchManagedCollisionModule` module, since exiting benchmark workflow only takes several minutes on the subset of the criteo-kaggle dataset. By default the identities become evictable after one hour. This descrepency leads to non-eviction during the profiling process. ### Dataset - [Criteo Kaggle Small](https://drive.google.com/file/d/1__rPcUSa45FHkmnBwivuM7K4nMYWD7b7/view?usp=sharing) - [Criteo Kaggle](https://drive.google.com/file/d/1_lAbXTEOk5vlPGXd4UvTrxGV6sCPer_R/view?usp=drive_link) Differential Revision: D76150895
|
This pull request was exported from Phabricator. Differential Revision: D76150895 |
lizhouyu
force-pushed
the
export-D76150895
branch
from
8036d23 to
2de3bb3
Compare
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.
Summary:
Benchmark framework for MPZCH
Rollback Plan:
Differential Revision: D76150895