Skip to content

Commit

Permalink
fix qat docs (PaddlePaddle#971)
Browse files Browse the repository at this point in the history
  • Loading branch information
yghstill authored Jan 28, 2022
1 parent fd13584 commit 7b49bed
Show file tree
Hide file tree
Showing 2 changed files with 4 additions and 4 deletions.
4 changes: 2 additions & 2 deletions demo/dygraph/quant/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -41,9 +41,9 @@ quant_config = {

- `'activation_preprocess_type'`:代表对量化模型激活值预处理的方法,目前支持PACT方法,如需使用可以改为'PACT';默认为None,代表不对激活值进行任何预处理。

- `weight_quantize_type`:代表模型权重的量化方式,可选的有['abs_max', 'moving_average_abs_max', 'channel_wise_abs_max'],默认为channel_wise_abs_max
- `weight_quantize_type`:代表模型权重的量化方式,可选的有['abs_max', 'channel_wise_abs_max'],默认为channel_wise_abs_max

- `activation_quantize_type`:代表模型激活值的量化方式,可选的有['abs_max', 'moving_average_abs_max'],默认为moving_average_abs_max
- `activation_quantize_type`:代表模型激活值的量化方式,可选的有['moving_average_abs_max'],默认为moving_average_abs_max

- `quantizable_layer_type`:代表量化OP的类型,目前支持Conv2D和Linear

Expand Down
4 changes: 2 additions & 2 deletions docs/zh_cn/api_cn/dygraph/quanter/qat.rst
Original file line number Diff line number Diff line change
Expand Up @@ -63,10 +63,10 @@ QAT
# activation预处理方法,默认为None,代表不进行预处理`
'activation_preprocess_type': None,
# weight量化方法, 默认为'channel_wise_abs_max', 此外还支持'channel_wise_abs_max'
# weight量化方法, 默认为'channel_wise_abs_max', 此外还支持'abs_max'
'weight_quantize_type': 'channel_wise_abs_max',
# activation量化方法, 默认为'moving_average_abs_max', 此外还支持'abs_max'
# activation量化方法, 默认为'moving_average_abs_max'
'activation_quantize_type': 'moving_average_abs_max',
# weight量化比特数, 默认为 8
Expand Down

0 comments on commit 7b49bed

Please sign in to comment.