Ebenezer Scrooge

作者您好，小弟在阅读论文时记得特征输入是上面Xdoc下面为0矩阵的一整个矩阵，但是看代码上： ``` input_ids, attention_mask = encode_input(text, model.tokenizer) input_ids = th.cat([input_ids[:-nb_test], th.zeros((nb_word, max_length), dtype=th.long), input_ids[-nb_test:]]) ``` 似乎零矩阵的下面，还有一部分从BERT做文档嵌入后的结果，这是为何呢？感谢！

Hello I am a player focusing on stealthy of jet fighters, recently I heard about ion stealthy tech in real life so I wanna make the ion thruster of origin...

``` # input: encoded sentence vector # output: encoded sentence vector in dataset with highest cosine similarity def find_similar_encoding(sent_vect): all_cosine = [] for sent in sent_encoded: result = 1 -...

Since TFLearn is used in model (AlexNet) constrction, thus tf.contrib (which is removed in TF 2.0) is necessary for this repo, plz add version requirement for TF in README.MD. Thx

win10 py3.7 with gdb 8.0.1 meliae can be build by neither pip (unistd.h : no such file or directory) nor setup.py (with print replaced by print())

如题： 另外，发现io.h及相关INCLUDE需要安装Win10 SDK，最后发现是Python 3.7 _ts结构体发生变化，不再含有exc_type，导致bbox.c相关代码编译失败

I see states and events are registered via annotations, which is hard to be altered during runtime, I wonder if it can generate FSM from given variables? Thx!

win10 py3.7 pip install failed

1

meliae\_scanner.c(600): fatal error C1083: 无法打开包括文件: “unistd.h”: No such file or directory