Implementation of nn.quantized.linear module #21921
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
Differential Revision: D15375695 Differential Version: 85144759
pytorchbot
added
module: nn
z-a-f
reviewed
Jun 18, 2019
Differential Revision: D15375695 Differential Version: 85155178
Differential Revision: D15375695 Differential Version: 85219380
dzhulgakov
reviewed
Jun 20, 2019
Differential Revision: D15375695 Differential Version: 85439410
dzhulgakov
reviewed
Jun 24, 2019
jerryzh168
commented
Jun 24, 2019
Differential Revision: D15375695 Differential Version: 85477647
Differential Revision: D15375695 Differential Version: 85509305
Differential Revision: D15375695 Differential Version: 85588772
Differential Revision: D15375695 Differential Version: 85589881
dzhulgakov
approved these changes
Jun 26, 2019
Differential Revision: D15375695 Differential Version: 85634127
raghuramank100
suggested changes
Jun 26, 2019
torch/nn/quantized/modules/linear.py
Outdated
| __constants__ = ['bias', 'in_features', 'out_features'] | ||
|
|
||
| def __init__(self, in_features, out_features, bias=True): | ||
| assert bias, 'nobias is not supported Quantized Linear module yet' |
There was a problem hiding this comment.
We should handle case with bias=False. In this case, there will be no self.bias to delete. We would still create qbias and set it to all zeros.
There was a problem hiding this comment.
you mean just accept the argument and keep the behavior the same?
There was a problem hiding this comment.
if we use this hack then quantized module will always have bias after deserialization even when user specify bias=False.
I think we can only support it if fbgemm has nobias implementation
Differential Revision: D15375695 Differential Version: 85655452
Differential Revision: D15375695 Differential Version: 85656064
raghuramank100
approved these changes
Jun 27, 2019
Differential Revision: D15375695 Differential Version: 85723371
|
This pull request has been merged in 5c46e70. |
xzhu1900
pushed a commit
to xzhu1900/pytorch
that referenced
this pull request
Jul 5, 2019
Summary: Pull Request resolved: pytorch#21921 Call FBGEMM kernels to implement quantized linear operator. This operator is used only for inference. Differential Revision: D15375695 fbshipit-source-id: b9ca6c156fd60481fea83e55603b2897f7bfc3eb
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:
:white_circle: #21930 nn.quantized.Relu and nn.quantize.Quantize/DeQuantize modules 💛
:white_circle: #21925 [quant] Add serialization for nn.quantized.Linear module 💛
:black_circle: #21921 Implementation of nn.quantized.linear module 💚
Call FBGEMM kernels to implement quantized linear operator. This operator is used only for inference.
Differential Revision: D15375695