This repository has been archived by the owner on Sep 18, 2024. It is now read-only.
Add bias quantization in QAT and refactor the code of weight quantization #2914
Conversation
This file contains 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
linbinskn
changed the title
QuanluZhang
reviewed
Sep 28, 2020
QuanluZhang
reviewed
Sep 28, 2020
| out[out == 0] = 1 | ||
| return out | ||
| weight[weight == 0] = 1 | ||
| wrapper.module.weight.data = weight |
There was a problem hiding this comment.
return is removed? returned value is never used?
There was a problem hiding this comment.
Returned value is quantized weight. There is no need to return in this version because weight is quantized in place in the function quantize_weight().
chicm-ms
reviewed
Sep 30, 2020
chicm-ms
reviewed
Oct 7, 2020
| assert math.isclose(model.conv2.module.scale, 6 / 255, abs_tol=eps) | ||
| assert model.conv2.module.zero_point in (42, 43) | ||
| # test value of weight and bias after quantization | ||
| weight = torch.tensor([[1.1287, 2.3456], [3.7814, 5.9723]]) | ||
| weight_valid = torch.tensor([[1.1242, 2.3421], [3.7707, 5.9723]]) |
There was a problem hiding this comment.
just curious, how are the values of weight_valid and bias_valid calculated?
There was a problem hiding this comment.
Weight_valid and bias_valid are calculated by quantization function manually. I will modify the test case and annotation after code freeze.
QuanluZhang
reviewed
Oct 10, 2020
| weight_bits = get_bits_length(config, 'weight') | ||
| quant_start_step = config.get('quant_start_step', 0) | ||
| assert weight_bits >= 1, "quant bits length should be at least 1" | ||
|
|
||
| if quant_start_step > self.steps: | ||
| return weight | ||
| return |
There was a problem hiding this comment.
Ok, I have fixed it.
QuanluZhang
reviewed
Oct 10, 2020
QuanluZhang
approved these changes
Oct 10, 2020
chicm-ms
approved these changes
Oct 10, 2020
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
Add bias quantization and refactor the code. If this PR was merged, I would submit a new PR to modify the doc.