renmada

[论文复现]如何创建复数的parameter

3

### 问题描述 Please describe your issue 1. paddle.create_parameter不能创建complex64的类型 2. set value 报错 ``` python self.weight = paddle.create_parameter([self.num_freqs, in_channels, out_channels], dtype=paddle.float32) self.weight.set_value(paddle.as_complex(paddle.randn([self.num_freqs, in_channels, out_channels, 2]))) ``` ``` need tensor with dtype...

self.init_weights()是重新初始化了bert的权重吗？

2

如果是的话，为什么要这么做

自动压缩后，用paddlelite 导出报错

2

测试自动压缩[demo](https://github.com/PaddlePaddle/PaddleSlim/tree/develop/demo/auto_compression)后，用paddlelite导出后报错 [F 5/21 18: 1: 6.617 ...optimizer/mir/static_kernel_pick_pass.cc:71 Apply] Check failed: !instruct.kernels().empty(): No kernels found for fake_quantize_abs_max

https://github.com/PaddlePaddle/models/tree/develop/PaddleCV/image_classification

No module named 'textclf.utils'

2

pip装的 (base) vocust001@ubuntu:~/anaconda3/lib/python3.7/site-packages$ textclf Traceback (most recent call last): File "/home/vocust001/anaconda3/bin/textclf", line 11, in load_entry_point('textclf==0.1.0', 'console_scripts', 'textclf')() File "/home/vocust001/anaconda3/lib/python3.7/site-packages/pkg_resources/__init__.py", line 490, in load_entry_point return get_distribution(dist).load_entry_point(group, name) File "/home/vocust001/anaconda3/lib/python3.7/site-packages/pkg_resources/__init__.py", line 2859,...

label编码的代码在哪

1

dataset里的代码是 if self.training: if not self.case_sensitive: label = label.lower() return (img, label) DistCollateFn里没有看到相关代码 输入给模型的label不能是str吧

在看代码里，发现有ImageRotaryEmbed

论文复现 61：VoVNet

2

权重下载 https://pan.baidu.com/s/1s7EAQiRnbPnWDRjA1VgPfw?pwd=w1xt

ErnieForGeneration 多卡训练 paddle.load模型报错：“EOFError: Ran out of input”

7

模型是paddlenlp的ErnieForGeneration，用model.save_pretrianed保存的

rflearning 怎么处理大写字母？

5

词典用Scene_text_36.txt，不包含大写字母