-
Notifications
You must be signed in to change notification settings - Fork 742
Replace debug handle with from_node to trace operator transformation
#11532
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11532
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit fe54b74 with merge base 67b6009 ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D76168997 |
|
@pytorchbot label "release notes: none" |
|
This pull request was exported from Phabricator. Differential Revision: D76168997 |
c685371 to
8a43fe7
Compare
pytorch#11532) Summary: Pull Request resolved: pytorch#11532 X-link: pytorch/ao#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
pytorch#2339) Summary: X-link: pytorch/executorch#11532 Pull Request resolved: pytorch#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
pytorch#2339) Summary: X-link: pytorch/executorch#11532 Pull Request resolved: pytorch#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
pytorch#2339) Summary: X-link: pytorch/executorch#11532 Pull Request resolved: pytorch#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
|
This pull request was exported from Phabricator. Differential Revision: D76168997 |
8a43fe7 to
89503bf
Compare
pytorch#11532) Summary: Pull Request resolved: pytorch#11532 X-link: pytorch/ao#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
|
This pull request was exported from Phabricator. Differential Revision: D76168997 |
pytorch#11532) Summary: Pull Request resolved: pytorch#11532 X-link: pytorch/ao#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
89503bf to
d3093f9
Compare
pytorch#11532) Summary: Pull Request resolved: pytorch#11532 X-link: pytorch/ao#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
|
This pull request was exported from Phabricator. Differential Revision: D76168997 |
d3093f9 to
fe54b74
Compare
pytorch#2339) Summary: X-link: pytorch/executorch#11532 Pull Request resolved: pytorch#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
#2339) Replace debug handle with `from_node` to trace operator transformation (#2339) Summary: X-link: pytorch/executorch#11532 Pull Request resolved: #2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
pytorch#2339) Replace debug handle with `from_node` to trace operator transformation (pytorch#2339) Summary: X-link: pytorch/executorch#11532 Pull Request resolved: pytorch#2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
| nonlocal prev_decomp_op_to_debug_handle_map | ||
| if FROM_NODE_KEY in node.meta: | ||
| prev_decomp_op = str(node.meta.get("nn_module_stack")) | ||
| debug_handle = _generate_debug_handle_from_node(node) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
is from_node a new thing?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
it is introduced after torch 2.6 and we will use that to trace the node transformation from torch.export to torchao to edge.
digantdesai
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is mainly updating the test, so ok with merging this.
#2339) Replace debug handle with `from_node` to trace operator transformation (#2339) Summary: X-link: pytorch/executorch#11532 Pull Request resolved: #2339 This diff replace the debug handle with `from_node` infrastructure, which is a first class citizen in exported program and used to trace the node-level transformation by recording every ancestor of given node. N6213836 is a demonstration of how `from_node` infra records the node transformation after unlifting and re-exporting exported graph. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle with hashing their greatest ancestor's node. After this change user no longer need to invoke `generate_numeric_debug_handle` for debugging. Also the original pipeline will still work under current scenario. Reviewed By: jerryzh168 Differential Revision: D76168997
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
Summary:
X-link: pytorch/ao#2339
This diff replace the debug handle with
from_nodeinfrastructure, which is a first class citizen in exported program and used to trace the node-level transformation. For simplify the progress, we are trying to reuse the debug handle infrastructure by generating debug handle from from_node info via hasing.After this change user no longer need to invoke
generate_numeric_debug_handlefor debugging. Also the original pipeline will still work under current scenario.Reviewed By: jerryzh168
Differential Revision: D76168997