RT.

RT.

当前开源出来的模型是否支持infilling 代码注释生成，代码翻译，代码翻译等功能呢？

基于公司数据做持续预训练, 可否提供预训练脚本,包括 lr，max_seq_len, 及其他需要注意的细节等

在源码中 input_wrapper 可以很好的包装代码生成prompt。

1.请问普通问答模式的prompt应该怎么包装？像chatgtp那样问答
例如：
Q：请使用C#生成一个快速排序算法
A：xxxx

2.发现一个问题，就是代码生成完成之后的末尾，最后都会带上一个

买的 ecs，基本没有cuda，能支持cpu运行就好了

We can record all the command that was hit during an activity using script functionality in linux and can use this code generator to create a script to perform that activity

-- All the parameters that are used during activity can be asked at one go in form of a batch file

Used Chatgpt to explain more

In our current workflow, documenting and replicating complex activities in Linux environments is time-consuming and prone to errors. Manually recording every command executed during a task is tedious and often leads to inconsistencies. Moreover, recreating these activities requires manual intervention and may result in deviations from the original process. This inefficiency not only hampers productivity but also poses risks to the reliability and stability of our operations.

Objective:
The primary objective of this project is to develop a tool that automates the recording of commands executed during activities in Linux environments. This tool will capture the command sequence and generate a script that can be used to replicate the activity accurately. Furthermore, the tool will incorporate functionality to create batch files that prompt users for parameters, simplifying the execution of tasks with varying inputs.

Solution Overview:
The proposed solution consists of two main components:

Command Recording Module: This module will intercept and record all commands executed within a designated session or timeframe. It will capture the command sequence along with relevant metadata such as timestamps and user identifiers. The recorded data will be stored in a structured format for further processing. -- > This part is already done using "script" command in linux

Script Generation Module: Upon completion of an activity, the recorded commands will be processed by the script generation module to create a reproducible script. This script will encapsulate the sequence of commands required to perform the activity, ensuring consistency and accuracy in subsequent executions. Additionally, the module will provide an option to generate batch files that prompt users for input parameters, enhancing flexibility and usability.

Benefits:

Time Savings: By automating the process of recording and scripting activities, we can significantly reduce the time required to document and replicate tasks.
Accuracy and Consistency: The generated scripts ensure that activities are performed consistently, reducing the risk of errors and deviations.
Usability: Batch file generation simplifies the execution of tasks by prompting users for input parameters, making it easier to adapt scripts to different scenarios.
Knowledge Sharing and Collaboration: Standardized scripts enable seamless sharing of best practices and facilitate collaboration among team members.
Audit Trail: The recorded command history provides a detailed audit trail of activities, enhancing accountability and compliance.

你好，请问下，code_string和later_code的输入长度有限制吗？

Congratulations on this wonderful work! I noticed that the Evol-Instruct method is utilized in aiXcoder-7B training. There are some differences between the traditional implementation of Evol-Instruct and aiXcoder's prompts modified based on FIM. Is there any specific implementation strategy or example for it? Thanks!

I have deployed an aiXcoder model locally, but when using the aiXcoder plugin, I noticed that it requires me to log in. After a successful login, I couldn't find an option to configure the plugin to point to my self-hosted aiXcoder model. Is there a way to set the plugin to use a locally deployed model? Or are there any plans to introduce this configuration option in the future? Many thanks!

Thanks for the excellent work. Waiting for the update!

看测试结果，aiXcoder-7B的准确率更高

RT

sess_megatron.py 这个类中，如果要改成可多并发而且是流式输出的应该怎么写

看到def run_infer(）函数有一个while True: 循环收集数据

可以请教一下用什么工具实现语法检查和静态分析的吗，可否开源一些脚本，非常感谢

请问如果我要用aixcoder-7b做一个纯生成的任务,而不是FIM格式的,例如一个代码网页问答页面, 参数如何设置.
比如 aixcoder-7b在测试生成类题目,如human-eval python的时候, later_code和file_path,解码方式以及推理参数是如何设置的?

找了半天，没看到相关信息

如题，想问问各位大佬是否有gradio端界面调用aiXcoder-7B模型

作为一个新手，我在部署的时候遇到了OOM的问题，我想问一下怎么样才能够多卡部署，能够给我一个示例吗？我将万分感激！

本地部署怎么进行vscode的代码补全？有推荐的插件吗？应该使用7B-base模型还是7B模型？

其他代码辅助的大模型有这些功能，还可以自动生成API文档。

目前aixcoder7b支持使用vllm加速推理吗？

While previous version didn't explain how the weights should be loaded, most runs used the default loading method (downloaded from the modelscope), this version some instructions will be added to help those who want to run the model based on the hugging face framework loading local weight.

deepseek 7B 的EM结果平均只有10多分，但是santacoder的数据结果和report的差不多

我看了下你的 sess_huggingface.py文件，我的需求是想做一个代码补全的demo，但是发现回复的内容太多，我在思考是否跟input_wrapper提供的参数设置有关系
调用例子如以下：
"code_string": "The programming language I am using is Java. I only need you to complete the possible code that may be written at the end of my code, without providing any extra explanation or description. Please directly complete the code for your response field. If it cannot be processed, return an empty string. The current code is public static void main(String[] args) {\n SpringApplication application \u003d new SpringApplication(TranCoderApplication.class);\n application.addInitializers(SpringBeanUtil::setApplicationContext);\n spring."
我希望接口返回run(args);
但是实际接口返回：

the file path is: test.py

the code file is written by Python

The programming language I am using is Java. I only need you to complete the possible code that may be written at the end of my code, without providing any extra explanation or description. Please directly complete the code for your response field. If it cannot be processed, return an empty string. The current code is public static void main(String[] args) {
SpringApplication application = new SpringApplication(TranCoderApplication.class);
application.addInitializers(SpringBeanUtil::setApplicationContext);
spring.application.run(args);
}

Assistant:

public static void main(String[] args) {
SpringApplication application = new SpringApplication(TranCoderApplication.class);
application.addInitializers(SpringBeanUtil::setApplicationContext);
spring.application.run(args);
}