兰铃

Get error when use albumentations: It seems that scikit-image has not been built correctly.

6

- [x] : I have read the checklist. # bug description background: The server I am using is not connected to the Internet therefore cannot use `pip`, so, I packed...

I want to use SSL to my project, [LFW](http://vis-www.cs.umass.edu/lfw/) is only a demo. Model structure, loss function, optimizer is keeping the same as this repo, only one GPU, load ResNet50...

I found that `m_queue.size()` isn't equal to zero when `pool.shutdown()` was runned, so some tasks do not finish when the thread pool closes. I fixed it up as follows: ```cpp...

Bug report: When sub-task are performed quickly, the results are inaccurate

4

```cpp #include #include "../include/ThreadPool.h" class TEST { private: std::mutex mtx; std::vector v1[2]; std::priority_queue q; public: void init() { v1[0].push_back(1); v1[0].push_back(2); v1[0].push_back(3); v1[1].push_back(1); v1[1].push_back(3); v1[1].push_back(2); } void push(const int& index) {...

Hello, thanks for you code, I'm als prepare to use SSL to improve robustness. Some question: https://github.com/Kim-Minseon/RoCL/blob/b6d5185e294e8bca5670e9146df81b54d99d0635/src/loss.py#L59 This loss function is from paper simCLR ? I also decided to use...

关于 spearman 得分很高但检索很差的询问

6

数据集是 0-1 格式的数据集，0 表示不相似，1 表示相似，各有10 万对。语料库中总共有 200 万个句子，也就是说有些样本没参与训练。不太一样的就是：第一个句子很短，5 6 个字左右，第二个句子很长，50字左右。比如淘宝搜索：XXX零食，推荐的结果会有：XXX商店XXX口味XXX面包。 我使用 CoSENT 进行训练，在 bert 后接入一个降维层，生成文本的 128 维度特征向量，期待相似样本的距离近，不相似样本距离远。微调 3 个 epoch 左右，spearman 得分在 0.86 左右。 而后，将所有文本生成特征向量，构建向量索引（这里用的别人成熟的框架，不是构建索引出错），并查询距离最近的向量，发现很难查回正样本，MRR 指标也很差，这个问题我怀疑是距离还是没有拉开，请问您在使用这个方法的时候有没有遇到过类似问题呢？~~仿佛偏题了~~

Make gui better and fix bug of QWebEngine doesn't display eq on linux

5

1. `app.exec_()` is pyqt4's method, `app.exec()` is pyqt5's method 2. I found that math equation displays well in windows but doesn't display in linux platform as description in #109 ,...

# 问题描述 虽然全部能AC过去，but，好像有个数算错了。 # 题目示例 **Sample In** `169 5 2` **Sample Out** `169 = 6^2 + 6^2 + 6^2 + 6^2 + 5^2` 我们来算一下：  答案正确。 # 实际代码 从github拷贝，g++编译。...

询问损失函数值

2

就是问一下，使用 cosent 这个损失函数，最后会收敛到多少呢？或者说收敛情况如何呢？我发现我这里训练半个 epoch 就很难继续去收敛了。 我的数据集是，20万对 0-1 样本，10万正，10万负。使用的 chinese-roberta-wwm-ext 微调，下游任务是 last hidden state 经过 max pool 、全连接和 normalization 后的 128 维度向量尽可能接近。 其他参数的话，16 bacth，2e-5 的学习率并 torch OneCycleLR 调整（类似 Warmup ），训练 5...

error: inlining failed in call to always_inline ‘__m128i _mm_alignr_epi8(__m128i, __m128i, int)’: target specific option mismatch

1

``` /usr/lib/gcc/x86_64-linux-gnu/7/include/tmmintrin.h:185:1: error: inlining failed in call to always_inline ‘__m128i _mm_alignr_epi8(__m128i, __m128i, int)’: target specific option mismatch _mm_alignr_epi8(__m128i __X, __m128i __Y, const int __N) ```