[quant][pyper] Support quantization of ops in fork-wait subgraph #44048
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
Summary:
Inline the fork-wait calls to make sure we can see the ops to be quantized in the main graph
Also fix the InlineForkWait JIT pass to account for the case where the aten::wait call isn't present in the main graph
and we return future tensor from subgraph
Example
```
graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_6325.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : Future[Tensor[]] = prim::fork_0(%self.1, %argument_1.1, %argument_2.1) # :0:0
return (%3)
with prim::fork_0 = graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_5396.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : __torch__.dper3.core.interop.___torch_mangle_6330.DperModuleWrapper = prim::GetAttr[name="x"](%self.1)
%4 : __torch__.dper3.core.interop.___torch_mangle_5397.DperModuleWrapper = prim::GetAttr[name="y"](%self.1)
%5 : __torch__.dper3.core.interop.___torch_mangle_6327.DperModuleWrapper = prim::GetAttr[name="z"](%4)
%6 : Tensor = prim::CallMethod[name="forward"](%5, %argument_1.1, %argument_2.1) # :0:0
%7 : None = prim::CallMethod[name="forward"](%3, %6) # :0:0
%8 : Tensor[] = prim::ListConstruct(%6)
return (%8)
```
Test Plan:
python test/test_quantization.py test_interface_with_fork
Reviewers:
Subscribers:
Tasks:
Tags:
[ghstack-poisoned]
This was referenced Sep 2, 2020
facebook-github-bot
added
the
oncall: jit
💊 CI failures summary and remediationsAs of commit b4423cd (more details on the Dr. CI page):
🕵️ 2 new failures recognized by patternsThe following CI failures do not appear to be due to upstream breakages:
|
…bgraph"
Summary:
Inline the fork-wait calls to make sure we can see the ops to be quantized in the main graph
Also fix the InlineForkWait JIT pass to account for the case where the aten::wait call isn't present in the main graph
and we return future tensor from subgraph
Example
```
graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_6325.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : Future[Tensor[]] = prim::fork_0(%self.1, %argument_1.1, %argument_2.1) # :0:0
return (%3)
with prim::fork_0 = graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_5396.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : __torch__.dper3.core.interop.___torch_mangle_6330.DperModuleWrapper = prim::GetAttr[name="x"](%self.1)
%4 : __torch__.dper3.core.interop.___torch_mangle_5397.DperModuleWrapper = prim::GetAttr[name="y"](%self.1)
%5 : __torch__.dper3.core.interop.___torch_mangle_6327.DperModuleWrapper = prim::GetAttr[name="z"](%4)
%6 : Tensor = prim::CallMethod[name="forward"](%5, %argument_1.1, %argument_2.1) # :0:0
%7 : None = prim::CallMethod[name="forward"](%3, %6) # :0:0
%8 : Tensor[] = prim::ListConstruct(%6)
return (%8)
```
Test Plan:
python test/test_quantization.py test_interface_with_fork
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D23481003](https://our.internmc.facebook.com/intern/diff/D23481003)
[ghstack-poisoned]
supriyar
added a commit
that referenced
this pull request
Sep 2, 2020
Summary:
Inline the fork-wait calls to make sure we can see the ops to be quantized in the main graph
Also fix the InlineForkWait JIT pass to account for the case where the aten::wait call isn't present in the main graph
and we return future tensor from subgraph
Example
```
graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_6325.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : Future[Tensor[]] = prim::fork_0(%self.1, %argument_1.1, %argument_2.1) # :0:0
return (%3)
with prim::fork_0 = graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_5396.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : __torch__.dper3.core.interop.___torch_mangle_6330.DperModuleWrapper = prim::GetAttr[name="x"](%self.1)
%4 : __torch__.dper3.core.interop.___torch_mangle_5397.DperModuleWrapper = prim::GetAttr[name="y"](%self.1)
%5 : __torch__.dper3.core.interop.___torch_mangle_6327.DperModuleWrapper = prim::GetAttr[name="z"](%4)
%6 : Tensor = prim::CallMethod[name="forward"](%5, %argument_1.1, %argument_2.1) # :0:0
%7 : None = prim::CallMethod[name="forward"](%3, %6) # :0:0
%8 : Tensor[] = prim::ListConstruct(%6)
return (%8)
```
Test Plan:
python test/test_quantization.py test_interface_with_fork
Reviewers:
Subscribers:
Tasks:
Tags:
ghstack-source-id: e2f0d12
Pull Request resolved: #44048
vkuzo
approved these changes
Sep 3, 2020
Contributor
vkuzo
left a comment
vkuzo
left a comment
There was a problem hiding this comment.
lg, although feel free to wait for @jerryzh168 if this needs a deeper review
…bgraph"
Summary:
Inline the fork-wait calls to make sure we can see the ops to be quantized in the main graph
Also fix the InlineForkWait JIT pass to account for the case where the aten::wait call isn't present in the main graph
and we return future tensor from subgraph
Example
```
graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_6325.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : Future[Tensor[]] = prim::fork_0(%self.1, %argument_1.1, %argument_2.1) # :0:0
return (%3)
with prim::fork_0 = graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_5396.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : __torch__.dper3.core.interop.___torch_mangle_6330.DperModuleWrapper = prim::GetAttr[name="x"](%self.1)
%4 : __torch__.dper3.core.interop.___torch_mangle_5397.DperModuleWrapper = prim::GetAttr[name="y"](%self.1)
%5 : __torch__.dper3.core.interop.___torch_mangle_6327.DperModuleWrapper = prim::GetAttr[name="z"](%4)
%6 : Tensor = prim::CallMethod[name="forward"](%5, %argument_1.1, %argument_2.1) # :0:0
%7 : None = prim::CallMethod[name="forward"](%3, %6) # :0:0
%8 : Tensor[] = prim::ListConstruct(%6)
return (%8)
```
Test Plan:
python test/test_quantization.py test_interface_with_fork
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D23481003](https://our.internmc.facebook.com/intern/diff/D23481003)
[ghstack-poisoned]
…bgraph"
Summary:
Inline the fork-wait calls to make sure we can see the ops to be quantized in the main graph
Also fix the InlineForkWait JIT pass to account for the case where the aten::wait call isn't present in the main graph
and we return future tensor from subgraph
Example
```
graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_6325.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : Future[Tensor[]] = prim::fork_0(%self.1, %argument_1.1, %argument_2.1) # :0:0
return (%3)
with prim::fork_0 = graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_5396.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : __torch__.dper3.core.interop.___torch_mangle_6330.DperModuleWrapper = prim::GetAttr[name="x"](%self.1)
%4 : __torch__.dper3.core.interop.___torch_mangle_5397.DperModuleWrapper = prim::GetAttr[name="y"](%self.1)
%5 : __torch__.dper3.core.interop.___torch_mangle_6327.DperModuleWrapper = prim::GetAttr[name="z"](%4)
%6 : Tensor = prim::CallMethod[name="forward"](%5, %argument_1.1, %argument_2.1) # :0:0
%7 : None = prim::CallMethod[name="forward"](%3, %6) # :0:0
%8 : Tensor[] = prim::ListConstruct(%6)
return (%8)
```
Test Plan:
python test/test_quantization.py test_interface_with_fork
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D23481003](https://our.internmc.facebook.com/intern/diff/D23481003)
[ghstack-poisoned]
supriyar
added a commit
that referenced
this pull request
Sep 4, 2020
Summary:
Inline the fork-wait calls to make sure we can see the ops to be quantized in the main graph
Also fix the InlineForkWait JIT pass to account for the case where the aten::wait call isn't present in the main graph
and we return future tensor from subgraph
Example
```
graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_6325.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : Future[Tensor[]] = prim::fork_0(%self.1, %argument_1.1, %argument_2.1) # :0:0
return (%3)
with prim::fork_0 = graph(%self.1 : __torch__.dper3.core.interop.___torch_mangle_5396.DperModuleWrapper,
%argument_1.1 : Tensor,
%argument_2.1 : Tensor):
%3 : __torch__.dper3.core.interop.___torch_mangle_6330.DperModuleWrapper = prim::GetAttr[name="x"](%self.1)
%4 : __torch__.dper3.core.interop.___torch_mangle_5397.DperModuleWrapper = prim::GetAttr[name="y"](%self.1)
%5 : __torch__.dper3.core.interop.___torch_mangle_6327.DperModuleWrapper = prim::GetAttr[name="z"](%4)
%6 : Tensor = prim::CallMethod[name="forward"](%5, %argument_1.1, %argument_2.1) # :0:0
%7 : None = prim::CallMethod[name="forward"](%3, %6) # :0:0
%8 : Tensor[] = prim::ListConstruct(%6)
return (%8)
```
Test Plan:
python test/test_quantization.py test_interface_with_fork
Reviewers:
Subscribers:
Tasks:
Tags:
ghstack-source-id: 8a919ff
Pull Request resolved: #44048
Contributor
|
This pull request has been merged in 199c73b. |
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:
Summary:
Inline the fork-wait calls to make sure we can see the ops to be quantized in the main graph
Also fix the InlineForkWait JIT pass to account for the case where the aten::wait call isn't present in the main graph
and we return future tensor from subgraph
Example
Test Plan:
python test/test_quantization.py test_interface_with_fork
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: D23481003