Unbreak test models llama CI #6026
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/6026
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 04b1dda with merge base b6e6d06 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D64074891 |
|
This pull request was exported from Phabricator. Differential Revision: D64074891 |
larryliu0820
added a commit
that referenced
this pull request
Oct 8, 2024
Summary: Pull Request resolved: #6026 Did a bunch of debugging on OSS CI:https://github.com/pytorch/executorch/actions/runs/11241297226/job/31252590975 Was able to confirm although the problem happens in `ConvertToLinear` but the root cause is we are partitioning the graph differently between these two pytorch nightly: dev20240916 and dev20240917. The exported graph looks the same but the partitioner was behaving differently and causes the `ConvertToLinear` pass to error out. We can't really revert back to dev20240916 nightly because it breaks other CI jobs, see #5987. The current approach I'm taking avoids decomposing linear by using `to_edge_lower_and_transform` API. This avoids jumping into the rabbit hole of debugging the partitioning & tagging logic. Differential Revision: D64074891
larryliu0820
force-pushed
the
export-D64074891
branch
from
91fba09 to
7e54dab
Compare
Jack-Khuu
approved these changes
Oct 8, 2024
facebook-github-bot
pushed a commit
that referenced
this pull request
Oct 8, 2024
Summary: Did a bunch of debugging on OSS CI:https://github.com/pytorch/executorch/actions/runs/11241297226/job/31252590975 Was able to confirm although the problem happens in `ConvertToLinear` but the root cause is we are partitioning the graph differently between these two pytorch nightly: dev20240916 and dev20240917. The exported graph looks the same but the partitioner was behaving differently and causes the `ConvertToLinear` pass to error out. We can't really revert back to dev20240916 nightly because it breaks other CI jobs, see #5987. The current approach I'm taking avoids decomposing linear by using `to_edge_lower_and_transform` API. This avoids jumping into the rabbit hole of debugging the partitioning & tagging logic. Reviewed By: Jack-Khuu Differential Revision: D64074891
facebook-github-bot
force-pushed
the
export-D64074891
branch
from
7e54dab to
bcc8b14
Compare
|
This pull request was exported from Phabricator. Differential Revision: D64074891 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D64074891 |
larryliu0820
added a commit
that referenced
this pull request
Oct 8, 2024
Summary: Pull Request resolved: #6026 Did a bunch of debugging on OSS CI:https://github.com/pytorch/executorch/actions/runs/11241297226/job/31252590975 Was able to confirm although the problem happens in `ConvertToLinear` but the root cause is we are partitioning the graph differently between these two pytorch nightly: dev20240916 and dev20240917. The exported graph looks the same but the partitioner was behaving differently and causes the `ConvertToLinear` pass to error out. We can't really revert back to dev20240916 nightly because it breaks other CI jobs, see #5987. The current approach I'm taking avoids decomposing linear by using `to_edge_lower_and_transform` API. This avoids jumping into the rabbit hole of debugging the partitioning & tagging logic. Reviewed By: Jack-Khuu Differential Revision: D64074891
larryliu0820
force-pushed
the
export-D64074891
branch
from
bcc8b14 to
0c2febd
Compare
facebook-github-bot
pushed a commit
that referenced
this pull request
Oct 8, 2024
Summary: Did a bunch of debugging on OSS CI:https://github.com/pytorch/executorch/actions/runs/11241297226/job/31252590975 Was able to confirm although the problem happens in `ConvertToLinear` but the root cause is we are partitioning the graph differently between these two pytorch nightly: dev20240916 and dev20240917. The exported graph looks the same but the partitioner was behaving differently and causes the `ConvertToLinear` pass to error out. We can't really revert back to dev20240916 nightly because it breaks other CI jobs, see #5987. The current approach I'm taking avoids decomposing linear by using `to_edge_lower_and_transform` API. This avoids jumping into the rabbit hole of debugging the partitioning & tagging logic. Reviewed By: Jack-Khuu Differential Revision: D64074891
facebook-github-bot
force-pushed
the
export-D64074891
branch
from
0c2febd to
70cc91f
Compare
|
This pull request was exported from Phabricator. Differential Revision: D64074891 |
tugsbayasgalan
approved these changes
Oct 9, 2024
Summary: Did a bunch of debugging on OSS CI:https://github.com/pytorch/executorch/actions/runs/11241297226/job/31252590975 Was able to confirm although the problem happens in `ConvertToLinear` but the root cause is we are partitioning the graph differently between these two pytorch nightly: dev20240916 and dev20240917. The exported graph looks the same but the partitioner was behaving differently and causes the `ConvertToLinear` pass to error out. We can't really revert back to dev20240916 nightly because it breaks other CI jobs, see #5987. The current approach I'm taking avoids decomposing linear by using `to_edge_lower_and_transform` API. This avoids jumping into the rabbit hole of debugging the partitioning & tagging logic. Reviewed By: digantdesai, Jack-Khuu, tugsbayasgalan Differential Revision: D64074891
facebook-github-bot
pushed a commit
that referenced
this pull request
Oct 9, 2024
Summary: Did a bunch of debugging on OSS CI:https://github.com/pytorch/executorch/actions/runs/11241297226/job/31252590975 Was able to confirm although the problem happens in `ConvertToLinear` but the root cause is we are partitioning the graph differently between these two pytorch nightly: dev20240916 and dev20240917. The exported graph looks the same but the partitioner was behaving differently and causes the `ConvertToLinear` pass to error out. We can't really revert back to dev20240916 nightly because it breaks other CI jobs, see #5987. The current approach I'm taking avoids decomposing linear by using `to_edge_lower_and_transform` API. This avoids jumping into the rabbit hole of debugging the partitioning & tagging logic. Reviewed By: digantdesai, Jack-Khuu, tugsbayasgalan Differential Revision: D64074891
facebook-github-bot
force-pushed
the
export-D64074891
branch
from
70cc91f to
75125be
Compare
larryliu0820
force-pushed
the
export-D64074891
branch
from
75125be to
04b1dda
Compare
|
This pull request was exported from Phabricator. Differential Revision: D64074891 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D64074891 |
|
This pull request has been merged in 72b3bb3. |
qma
added a commit
to qma/executorch
that referenced
this pull request
Oct 15, 2024
Summary: Since master has migrated aot_compiler to use to_edge_transform_and_lower in a previous change pytorch#6026, quantization XNNPack options can be enabled by default for the following models: - Quantized ViT - Quantized Mobilebert - Quantized Emformer Predict - Quantized Emformer Transcribe Differential Revision: D64081319
qma
added a commit
to qma/executorch
that referenced
this pull request
Oct 16, 2024
pytorch#6242) Summary: Since master has migrated aot_compiler to use to_edge_transform_and_lower in a previous change pytorch#6026, quantization XNNPack options can be enabled by default for the following models: - Quantized ViT - Quantized Mobilebert - Quantized Emformer Predict - Quantized Emformer Transcribe Reviewed By: digantdesai Differential Revision: D64081319
qma
added a commit
to qma/executorch
that referenced
this pull request
Oct 16, 2024
pytorch#6242) Summary: Since master has migrated aot_compiler to use to_edge_transform_and_lower in a previous change pytorch#6026, quantization XNNPack options can be enabled by default for the following models: - Quantized ViT - Quantized Mobilebert - Quantized Emformer Predict - Quantized Emformer Transcribe Reviewed By: digantdesai Differential Revision: D64081319
qma
added a commit
to qma/executorch
that referenced
this pull request
Oct 16, 2024
pytorch#6242) Summary: Since master has migrated aot_compiler to use to_edge_transform_and_lower in a previous change pytorch#6026, quantization XNNPack options can be enabled by default for the following models: - Quantized ViT - Quantized Mobilebert - Quantized Emformer Predict - Quantized Emformer Transcribe Reviewed By: digantdesai Differential Revision: D64081319
facebook-github-bot
pushed a commit
that referenced
this pull request
Oct 17, 2024
#6242) Summary: Pull Request resolved: #6242 Since master has migrated aot_compiler to use to_edge_transform_and_lower in a previous change #6026, quantization XNNPack options can be enabled by default for the following models: - Quantized ViT - Quantized Mobilebert - Quantized Emformer Predict - Quantized Emformer Transcribe Reviewed By: digantdesai Differential Revision: D64081319 fbshipit-source-id: 4e8ff77af442dfded043c5a5583466afec6beb4e
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:
Did a bunch of debugging on OSS CI:https://github.com/pytorch/executorch/actions/runs/11241297226/job/31252590975
Was able to confirm although the problem happens in
ConvertToLinearbut the root cause is we are partitioning the graph differently between these two pytorch nightly: dev20240916 and dev20240917.The exported graph looks the same but the partitioner was behaving differently and causes the
ConvertToLinearpass to error out.We can't really revert back to dev20240916 nightly because it breaks other CI jobs, see #5987.
The current approach I'm taking avoids decomposing linear by using
to_edge_lower_and_transformAPI. This avoids jumping into the rabbit hole of debugging the partitioning & tagging logic.Differential Revision: D64074891