ModelDeltaTracer implementation for tracking logic #3060
Closed
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: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 8, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
c838aac to
14aea87
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
force-pushed
the
export-D76094097
branch
from
14aea87 to
1bc7753
Compare
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 8, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 8, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
2 times, most recently
from
bc8bc81 to
a3b38b4
Compare
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
a3b38b4 to
5083f92
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
pushed a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
5083f92 to
6ee62dd
Compare
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
6ee62dd to
d81051a
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
d81051a to
860e5f2
Compare
aliafzal
pushed a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
860e5f2 to
5b2a529
Compare
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 9, 2025
Summary: ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 11, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
b0570d5 to
ab48286
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
force-pushed
the
export-D76094097
branch
from
ab48286 to
472b04c
Compare
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
472b04c to
80546fd
Compare
Summary: Pull Request resolved: pytorch#3059 # This Diff Added implementation for fqn_to_feature_names method along with initial testing framework and UTs for fqn_to_feature_names # ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D75908963 Reviewed By: kausv
aliafzal
pushed a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
80546fd to
fdf6c0d
Compare
aliafzal
pushed a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097 Reviewed By: TroyGarden
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
fdf6c0d to
3df7ce5
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
3df7ce5 to
aa54b2f
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
added a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
aliafzal
force-pushed
the
export-D76094097
branch
from
aa54b2f to
8cdde7f
Compare
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Reviewed By: TroyGarden Differential Revision: D76094097
|
This pull request was exported from Phabricator. Differential Revision: D76094097 |
aliafzal
force-pushed
the
export-D76094097
branch
from
8cdde7f to
9a81f6d
Compare
aliafzal
pushed a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097 Reviewed By: TroyGarden
aliafzal
pushed a commit
to aliafzal/torchrec
that referenced
this pull request
Jun 12, 2025
Summary: Pull Request resolved: pytorch#3060 ### Diff Summary This diff introduces implementation of tracking logic for ID and Embedding mode 1. **Record Functions** ```record_lookup():``` Handles recording of IDs and embeddings based on the tracking mode. ```record_ids():``` Records IDs from a KeyedJaggedTensor. ```record_embeddings():``` Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings. 2. **Delta Retrieval** ```get_delta():``` Retrieves per FQN local IDs for each sparse feature. 3. **Tracked Modules Access** ```get_tracked_modules():``` Returns a dictionary of tracked modules. ## ModelDeltaTracker Context ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for: 1. Identifying which embedding rows were accessed during model execution 2. Retrieving the latest delta or unique rows for a model 3. Computing top-k changed embeddings 4. Supporting streaming updated embeddings between systems during online training Differential Revision: D76094097 Reviewed By: TroyGarden
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:
Diff Summary
This diff introduces implementation of tracking logic for ID and Embedding mode
Record Functions
record_lookup():Handles recording of IDs and embeddings based on the tracking mode.record_ids():Records IDs from a KeyedJaggedTensor.record_embeddings():Records IDs along with embeddings, ensuring size compatibility between IDs and embeddings.Delta Retrieval
get_delta():Retrieves per FQN local IDs for each sparse feature.Tracked Modules Access
get_tracked_modules():Returns a dictionary of tracked modules.ModelDeltaTracker Context
ModelDeltaTracker is a utility for tracking and retrieving unique IDs and their corresponding embeddings or states from embedding modules in model using Torchrec. It's particularly useful for:
Differential Revision: D76094097