[quant][fx] Fix dynamic weighted op lowering when input is used multiple times #74364
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
…ple times
Summary:
if a input is used multiple times in modules that are dynamically quantized:
```
x -- linear1
\-- linear2
```
we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass
to duplicate dequantize ops for pattern matching:
```
x - quantize_per_tensor_dynamic - dequantize1 - linear1
\----- dequantize2 - linear2
```
But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes
we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case
to recover both patterns:
```
x - quantize_per_tensor_dynamic1 -- dequantize1 -- linear1
\- quantize_per-tensor_dynamic2 -- dequantize2 -- linear2
```
so that they can be fused into dynamic linear:
```
x - linear_dynamic1
\-- linear_dynamic2
```
Test Plan:
python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use
Reviewers:
Subscribers:
Tasks:
Tags:
[ghstack-poisoned]
CI Flow Status⚛️ CI FlowRuleset - Version:
|
Contributor
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit 27a8e0d (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
This was referenced Mar 17, 2022
Contributor
Author
|
@jerryzh168 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
… used multiple times"
Summary:
if a input is used multiple times in modules that are dynamically quantized:
```
x -- linear1
\-- linear2
```
we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass
to duplicate dequantize ops for pattern matching:
```
x - quantize_per_tensor_dynamic - dequantize1 - linear1
\----- dequantize2 - linear2
```
But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes
we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case
to recover both patterns:
```
x - quantize_per_tensor_dynamic1 -- dequantize1 -- linear1
\- quantize_per-tensor_dynamic2 -- dequantize2 -- linear2
```
so that they can be fused into dynamic linear:
```
x - linear_dynamic1
\-- linear_dynamic2
```
Test Plan:
python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D34952755](https://our.internmc.facebook.com/intern/diff/D34952755)
[ghstack-poisoned]
… used multiple times"
Summary:
if a input is used multiple times in modules that are dynamically quantized:
```
x -- linear1
\-- linear2
```
we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass
to duplicate dequantize ops for pattern matching:
```
x - quantize_per_tensor_dynamic - dequantize1 - linear1
\----- dequantize2 - linear2
```
But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes
we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case
to recover both patterns:
```
x - quantize_per_tensor_dynamic1 -- dequantize1 -- linear1
\- quantize_per-tensor_dynamic2 -- dequantize2 -- linear2
```
so that they can be fused into dynamic linear:
```
x - linear_dynamic1
\-- linear_dynamic2
```
Test Plan:
python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D34952755](https://our.internmc.facebook.com/intern/diff/D34952755)
[ghstack-poisoned]
Contributor
Author
|
@jerryzh168 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
andrewor14
approved these changes
Mar 17, 2022
| # run the weight observer | ||
| weight_observer_module() | ||
|
|
||
| # this method is temporary will be removed soon |
Contributor
There was a problem hiding this comment.
Is the real fix removing the check for whether dequantize nodes have multiple users in the lowering code? Is there a reason why we don't just directly change that instead of adding a temporary fix?
Contributor
Author
There was a problem hiding this comment.
yes, the real fix is to stop duplicating dequantize and remove the check, planning to talk to you in today's sync
… used multiple times"
Summary:
if a input is used multiple times in modules that are dynamically quantized:
```
x -- linear1
\-- linear2
```
we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass
to duplicate dequantize ops for pattern matching:
```
x - quantize_per_tensor_dynamic - dequantize1 - linear1
\----- dequantize2 - linear2
```
But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes
we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case
to recover both patterns:
```
x - quantize_per_tensor_dynamic1 -- dequantize1 -- linear1
\- quantize_per-tensor_dynamic2 -- dequantize2 -- linear2
```
so that they can be fused into dynamic linear:
```
x - linear_dynamic1
\-- linear_dynamic2
```
Test Plan:
python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D34952755](https://our.internmc.facebook.com/intern/diff/D34952755)
[ghstack-poisoned]
… used multiple times"
Summary:
if a input is used multiple times in modules that are dynamically quantized:
```
x -- linear1
\-- linear2
```
we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass
to duplicate dequantize ops for pattern matching:
```
x - quantize_per_tensor_dynamic - dequantize1 - linear1
\----- dequantize2 - linear2
```
But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes
we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case
to recover both patterns:
```
x - quantize_per_tensor_dynamic1 -- dequantize1 -- linear1
\- quantize_per-tensor_dynamic2 -- dequantize2 -- linear2
```
so that they can be fused into dynamic linear:
```
x - linear_dynamic1
\-- linear_dynamic2
```
Test Plan:
python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D34952755](https://our.internmc.facebook.com/intern/diff/D34952755)
[ghstack-poisoned]
Contributor
Author
|
@jerryzh168 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
jerryzh168
added
release notes: quantization
facebook-github-bot
pushed a commit
that referenced
this pull request
Mar 18, 2022
…ple times (#74364) Summary: Pull Request resolved: #74364 if a input is used multiple times in modules that are dynamically quantized: ``` x -- linear1 \-- linear2 ``` we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass to duplicate dequantize ops for pattern matching: ``` x - quantize_per_tensor_dynamic - dequantize1 - linear1 \----- dequantize2 - linear2 ``` But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case to recover both patterns: ``` x - quantize_per_tensor_dynamic1 -- dequantize1 -- linear1 \- quantize_per-tensor_dynamic2 -- dequantize2 -- linear2 ``` so that they can be fused into dynamic linear: ``` x - linear_dynamic1 \-- linear_dynamic2 ``` Test Plan: python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use Imported from OSS Reviewed By: yixin94 Differential Revision: D34952755 fbshipit-source-id: a950159fd6a661e84faf0baf1692f6783904cfb3
shahofblah
pushed a commit
that referenced
this pull request
Mar 25, 2022
…ple times (#74364) Summary: Pull Request resolved: #74364 if a input is used multiple times in modules that are dynamically quantized: ``` x -- linear1 \-- linear2 ``` we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass to duplicate dequantize ops for pattern matching: ``` x - quantize_per_tensor_dynamic - dequantize1 - linear1 \----- dequantize2 - linear2 ``` But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case to recover both patterns: ``` x - quantize_per_tensor_dynamic1 -- dequantize1 -- linear1 \- quantize_per-tensor_dynamic2 -- dequantize2 -- linear2 ``` so that they can be fused into dynamic linear: ``` x - linear_dynamic1 \-- linear_dynamic2 ``` Test Plan: python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use Imported from OSS Reviewed By: yixin94 Differential Revision: D34952755 fbshipit-source-id: a950159fd6a661e84faf0baf1692f6783904cfb3 (cherry picked from commit 8a68968)
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.
Stack from ghstack (oldest at bottom):
Summary:
if a input is used multiple times in modules that are dynamically quantized:
we'll insert quantize_per_tensor_dynamic and dequantize for input, and we'll have a duplicate pass
to duplicate dequantize ops for pattern matching:
But we also have a check in the lowering code that if quantize_per_tensor_dynamic is used by multiple nodes
we'll skip the pattern, so the pattern is not recognized, we need to duplicate quantize_per_tensor_dynamic as well in this case
to recover both patterns:
so that they can be fused into dynamic linear:
Test Plan:
python test/test_quantization.py TestQuantizeFx.test_dynamic_linear_input_multiple_use
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: D34952755