modify tracin self influence helpers #994
Closed
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
|
This pull request was exported from Phabricator. Differential Revision: D35603078 |
|
cc: @99warriors it looks like some of the tests related to progress bar are failing. Do you mind looking into it ? |
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 22, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Reviewed By: NarineK Differential Revision: D35603078 fbshipit-source-id: 56efa7ca82253a71c3ea143f3e2f1cabbe483b58
99warriors
force-pushed
the
export-D35603078
branch
from
5e8177d to
d1d78d2
Compare
|
This pull request was exported from Phabricator. Differential Revision: D35603078 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D35603078 |
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 22, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Reviewed By: NarineK Differential Revision: D35603078 fbshipit-source-id: 8a84e1ce98be36b8eb0ac82c504417bdf21a6539
99warriors
force-pushed
the
export-D35603078
branch
from
d1d78d2 to
7e068c2
Compare
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 23, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Reviewed By: NarineK Differential Revision: D35603078 fbshipit-source-id: 8ce8db910a48b92b07efc4de1ec100d2939e4794
99warriors
force-pushed
the
export-D35603078
branch
from
7e068c2 to
0e0cbe2
Compare
|
This pull request was exported from Phabricator. Differential Revision: D35603078 |
NarineK
reviewed
Jul 25, 2022
| @@ -95,7 +96,7 @@ class TracInCPBase(DataInfluence): | |||
| def __init__( | |||
| self, | |||
| model: Module, | |||
| influence_src_dataset: Union[Dataset, DataLoader], | |||
| train_dataset: Union[Dataset, DataLoader], | |||
There was a problem hiding this comment.
@99warriors, do you mind fixing this naming also in the tutorials ? I remember that we used them explicitly as argument names.
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 26, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Reviewed By: NarineK Differential Revision: D35603078 fbshipit-source-id: 8944c0b909a2c6daa589b94e3df22d5f9e651346
99warriors
force-pushed
the
export-D35603078
branch
from
0e0cbe2 to
c96a460
Compare
|
This pull request was exported from Phabricator. Differential Revision: D35603078 |
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 29, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Reviewed By: NarineK Differential Revision: D35603078 fbshipit-source-id: 5f2164d1ebe75359c39b5a540a13eee6ab582900
99warriors
force-pushed
the
export-D35603078
branch
from
c96a460 to
535c7b4
Compare
|
This pull request was exported from Phabricator. Differential Revision: D35603078 |
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 29, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Differential Revision: D35603078 fbshipit-source-id: cd7ac1e59ffa5f57cfd9990626c44d7033636b7b
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 31, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Differential Revision: D35603078 fbshipit-source-id: 5c89f2f8ba5350c0c38e0cad6ccfd0929133c386
99warriors
added a commit
to 99warriors/captum
that referenced
this pull request
Jul 31, 2022
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Differential Revision: D35603078 fbshipit-source-id: 87063052e68441b82514489f4d9f9ad29b396da4
Summary: Pull Request resolved: pytorch#994 change `TracInCP._self_influence_batch_tracincp` and `TracInCP._self_influence_batch_tracincp` `TracInCP._self_influence_batches_tracincp_fast` to be named `self_influence`, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence. The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by `influence` when running it in self influence mode. The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large. Minor change: for `influence_src_dataset` argument of all `__init__`'s, add description of what assumptions we make of the batches yielded by the dataloader. Reviewed By: NarineK Differential Revision: D35603078 fbshipit-source-id: 78d233a5da210424f3eed308ee563d3baeba4135
99warriors
force-pushed
the
export-D35603078
branch
from
535c7b4 to
4e647df
Compare
|
This pull request was exported from Phabricator. Differential Revision: D35603078 |
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:
change
TracInCP._self_influence_batch_tracincpandTracInCP._self_influence_batch_tracincpTracInCP._self_influence_batches_tracincp_fastto be namedself_influence, which is now public, and now accept a DataLoader yielding batches (as well as a single batch, as before). The modified helper function can be called by external functions to compute self influence.The helper itself is also changed to improve efficiency, by reducing the number of times checkpoints are loaded. The modified helper, despite being able to compute self influence scores for a dataloader yielding batches, still only loads each checkpoint once, per call. This is because the modified helper now has an outer iteration over checkpoints, and an inner iteration over batches (the order of iteration is reversed compared to before). This helper is called by
influencewhen running it in self influence mode.The reason we cannot just increase the batch size to reduce the number of checkpoint loadings is that for large models (precisely those for which loading checkpoints is expensive), the model takes up too much memory, so that the batch size cannot be too large.
Minor change: for
influence_src_datasetargument of all__init__'s, add description of what assumptions we make of the batches yielded by the dataloader.Reviewed By: NarineK
Differential Revision: D35603078