Paddle

Paddle copied to clipboard

[BUG FIX]Fix MetaTensor's bug when run infermeta

1

### PR types Bug fixes ### PR changes Others ### Describe 迁移了sum算子后，dsin模型在执行infermeta的时候出现了断言错误如下：  原因是在获取dim的时候，selectedrows的Tensor未拿到正确的dim，该pr进行了修复

YuanRisheng

Adapting device-specific Extra Attributes for the PHI kernel

1

### PR types New features ### PR changes Others ### Describe Adapting device-specific Extra Attributes for the PHI kernel

chenwhql

[cherrypick] Fix elementwise_sub sign reverse for mkldnn

1

### PR types Bug fixes ### PR changes OPs ### Describe reference #46049

zh794390558

【论文复现】反向对齐paddle模型梯度异常，梯度无法正常打印

1

### bug描述 Describe the Bug 反向对齐过程中，打印paddle的模型参数梯度发现前面的几层大量参数的梯度为None 定位到梯度异常的layer以后，使用hook打印梯度发现异常 torch 代码 ```python class MLM(nn.Module): ''' Architecture of MLM ''' def __init__(self, n_dim=512): super(MLM, self).__init__() self.MLM_SequenceModeling_mask = Transforme_Encoder(n_layers=2, n_position=256) self.MLM_SequenceModeling_WCL = Transforme_Encoder(n_layers=1, n_position=256)...

vax521

Hapi mnist accuracy drop with commit "Enable standalone executor for single-GPU training (#45913)"

1

### 问题描述 Please describe your issue Commit https://github.com/PaddlePaddle/Paddle/commit/fb37628530823ad79baa0120511669205d6eba3b seems to be responsible of accuracy drop in **hapi mnist** model #### application: paddle_paddle/logs_paddle_paddle/build/python/paddle/tests/hapi_mnist_bf16_static.py #### test run parameters: FLAGS_use_mkldnn=true python /logs_paddle_paddle/build/python/paddle/tests/hapi_mnist_bf16_static.py --bf16=False...

pmajchrzak

[cherry-pick] [BugFix] fixed a bug that deco_name can't be parsed corrected (#46297)

2

### PR types Others ### PR changes Others ### Describe Fixed a bug that deco_name can't be parsed corrected Regular expression is used now

feifei-111

### PR types Others ### PR changes Others ### Describe - Flake8 tracking issue: #46039 修复 `remove-tabs` 存量 python 代码（`'\t'` -> `' ' * 4`） 并修正部分格式问题 本 PR 仅针对 `python/paddle/fluid/tests/`...

SigureMo

[Sparse] Add a batch_norm kernel

1

### PR types Others ### PR changes OPs ### Describe 因为：当前batch_norm是通过前端调用dense的batch_norm，需要多调用一次values和sparse_coo_tensor，也同时引入了对应的反向操作。 所以：添加batch_norm_kernel和batch_norm_grad_kernel，后端调用dense的kernel避免引入values和sparse_coo_tensor的调用。

zkh2016

fix bug in convert call: tranform the static func. (#46278)

1

### PR types Bug fixes ### PR changes Others ### Describe fix bug in convert call: tranform the static func. (#46278)

2742195759

[API] added 'place' dummy interface for variable

1

### PR types Others ### PR changes Others ### Describe 'place' interface is added to variable which exists in dygraph mode tensor

feifei-111