• 我是否读过这篇论文？
• 我读过OSDI 2021的哪些论文？
• 哪些论文标题中包含“Distributed”？
• 2020年有多少篇论文是由Mosharaf Chowdhury教授合著的？

好吧，问问 reason。

工作原理

过滤并列出论文

$ reason
>> # Show all papers.
>> ls
+----------------------------------------------------------+----------------+---------+------+
|                           title                          |  first author  |  venue  | year |
+============================================================================================+
| Shadowtutor: Distributed Partial Distillation for Mobile | Jae-Won Chung  | ICPP    | 2020 |
| Video DNN Inference                                      |                |         |      |
|----------------------------------------------------------+----------------+---------+------|
| Efficient Memory Disaggregation with Infiniswap          | Juncheng Gu    | NSDI    | 2017 |
|----------------------------------------------------------+----------------+---------+------|
| Refurbish Your Training Data: Reusing Partially          | Gyewon Lee     | ATC     | 2021 |
| Augmented Samples for Faster Deep Neural Network         |                |         |      |
| Training                                                 |                |         |      |
|----------------------------------------------------------+----------------+---------+------|
| Finding Consensus Bugs in Etherium via Multi-transaction | Youngseok Yang | OSDI    | 2021 |
| Differential Fuzzing                                     |                |         |      |
|----------------------------------------------------------+----------------+---------+------|
| Tiresias: A GPU Cluster Manager for Distributed Deep     | Juncheng Gu    | NSDI    | 2019 |
| Learning                                                 |                |         |      |
|----------------------------------------------------------+----------------+---------+------|
| Nimble: Lightweight and Parallel GPU Task Scheduling for | Woosuk Kwon    | NeurIPS | 2020 |
| Deep Learning                                            |                |         |      |
+----------------------------------------------------------+----------------+---------+------+
>> # Filter by 'title'. All these are regexes!
>> ls 'Deep Learning$'
+------------------------------------------------------------+--------------+---------+------+
|                            title                           | first author |  venue  | year |
+============================================================================================+
| Tiresias: A GPU Cluster Manager for Distributed Deep       | Juncheng Gu  | NSDI    | 2019 |
| Learning                                                   |              |         |      |
|------------------------------------------------------------+--------------+---------+------|
| Nimble: Lightweight and Parallel GPU Task Scheduling for   | Woosuk Kwon  | NeurIPS | 2020 |
| Deep Learning                                              |              |         |      |
+------------------------------------------------------------+--------------+---------+------+
>> # You may set default filters with `cd`.
>> # BTW, `cd .`, `cd ..`, `cd -`, and `cd` are supported, too.
>> cd 'Deep Learning$'
>> pwd
title matches 'Deep Learning$'
>> # Default filter are automatically applied.
>> # Infiniswap (NSDI'17) is not shown, because its title doesn't match 'Deep Learning$'.
>> ls at NSDI
+------------------------------------------------------------+--------------+---------+------+
|                            title                           | first author |  venue  | year |
+============================================================================================+
| Tiresias: A GPU Cluster Manager for Distributed Deep       | Juncheng Gu  | NSDI    | 2019 |
| Learning                                                   |              |         |      |
+------------------------------------------------------------+--------------+---------+------+
>> # Delete Tiresias.
>> ls at NSDI | rm
Removed 1 paper.

导入新论文

>> # Import directly from arXiv and USENIX. This will also download paper PDFs.
>> curl https://arxiv.org/abs/2105.11367
+--------------------------------------------------------+--------------+-------+------+
|                          title                         | first author | venue | year |
+======================================================================================+
| FedScale: Benchmarking Model and System Performance of | Fan Lai      | arXiv | 2021 |
| Federated Learning                                     |              |       |      |
+--------------------------------------------------------+--------------+-------+------+
>> curl https://www.usenix.org/conference/nsdi21/presentation/you
+------------------------------------------+--------------+-------+------+
|                   title                  | first author | venue | year |
+========================================================================+
| Ship Compute or Ship Data? Why Not Both? | Jie You      | NSDI  | 2021 |
+------------------------------------------+--------------+-------+------+
>> # Modify paper metadata.
>> ls ship | set as Kayak
>> # Or, import manually.
>> touch 'Batch Normalization: Accelerating Deep Network Training by Reducing Internal
Covariate Shift' by 'Sergey Ioffe, Christian Szegedy' at ICML in 2015 as BN @ BatchNorm.pdf
+--------------------------------------------------------------+--------------+-------+------+
|                             title                            | first author | venue | year |
+============================================================================================+
| Batch Normalization: Accelerating Deep Network Training by   | Sergey Ioffe | ICML  | 2015 |
| Reducing Internal Covariate Shift                            |              |       |      |
+--------------------------------------------------------------+--------------+-------+------+

阅读，做笔记，创建书籍！

>> # Open with a PDF viewer (`open`) and edit markdown notes with your editor (`ed`).
>> ls 'Why Not Both' | open | ed
+------------------------------------------+--------------+-------+------+
|                   title                  | first author | venue | year |
+========================================================================+
| Ship Compute or Ship Data? Why Not Both? | Jie You      | NSDI  | 2021 |
+------------------------------------------+--------------+-------+------+
>> # Format your markdown notes into an HTML book, and open it in your browser.
>> ls 'Deep Learning' | printf
+------------------------------------------------------------+--------------+---------+------+
|                            title                           | first author |  venue  | year |
+============================================================================================+
| Tiresias: A GPU Cluster Manager for Distributed Deep       | Juncheng Gu  | NSDI    | 2019 |
| Learning                                                   |              |         |      |
|------------------------------------------------------------+--------------+---------+------|
| Nimble: Lightweight and Parallel GPU Task Scheduling for   | Woosuk Kwon  | NeurIPS | 2020 |
| Deep Learning                                              |              |         |      |
+------------------------------------------------------------+--------------+---------+------+

printf

printf 会根据您的笔记创建一个HTML书籍。

命令

调用 reason 将启动一个新的命令提示符。它接受类似Unix的命令，但这些命令将在您的论文库中的研究论文上工作。

当前工作

• ls 以表格形式过滤并打印论文。默认列是标题、第一作者(by1)、会议(at) 和年份(in)。
• cd 向默认过滤器集添加一个AND过滤器（启动时默认为空）。
• pwd 显示由 cd 设置的当前默认过滤器集。
• touch 在您的论文库中创建一个新条目。
• curl 从网络上导入论文，例如arXiv或usenix.org。如果可用，它还会下载论文PDF。实验上支持下载原始PDF URL并推断元数据字段。
• rm 从您的论文库中删除条目。
• set 设置论文属性，包括自定义标签，这些标签还可以用于在 ls 中为论文着色。
• printf 使用 mdbook 创建您笔记的HTML页面。
• open 使用您的PDF查看器打开论文（可配置，默认为zathura）。
• ed 打开您的编辑器（可配置，默认为vim），您可以在其中编辑您的笔记。
• wc 计算论文数量。
• man 加上一个命令将打印该命令的文档。
• exit 或 Ctrl-d 退出 reason。

尚未，但希望很快（非常欢迎贡献！）

• grep 返回包含您指定的查询字符串的笔记的论文列表。
• sort 根据指定的列对论文进行排序。
• stat 打印论文的元数据和笔记。
• top 打印论文库的摘要。

安装

您可以从发行版中获取二进制文件，或者运行 cargo install reason-shell.