Special attention should be paid to the last step of configuring a domestic proxy, otherwise packages cannot be downloaded in China.

Option 1: Install using Conda (Recommended; no root privileges required)

Since you often use the Python environment, installing Go using conda is the simplest and cleanest way, and it doesn’t require administrator (sudo) privileges.

Bash

# 在你的基础环境或特定环境中安装
conda install -c conda-forge go

Advantages: One-click installation, newer version, and does not contaminate the system environment.
Disadvantages: Only effective in the current Conda environment.

Option 2: Standard installation of Linux server (official binary)

If you need global installation (and have sudo permissions), or don’t want to rely on Conda, this is the standard approach.

Download installation package (using Linux x86-64 as an example, version 1.23.4):

Bash

wget https://go.dev/dl/go1.23.4.linux-amd64.tar.gz
# 如果 go.dev 访问慢，可以使用国内源：
# wget https://mirrors.ustc.edu.cn/golang/go1.23.4.linux-amd64.tar.gz

Unzip to the installation directory (usually /usr/local):

Bash

sudo rm -rf /usr/local/go && sudo tar -C /usr/local -xzf go1.23.4.linux-amd64.tar.gz

Configure environment variables:

The Go binary directory needs to be added to the PATH.

Bash

# 编辑 ~/.bashrc
echo 'export PATH=$PATH:/usr/local/go/bin' >> ~/.bashrc
# 同时也把 Go 安装第三方工具的目录加进去（比如安装 s5cmd 后会在 $HOME/go/bin）
echo 'export PATH=$PATH:$(go env GOPATH)/bin' >> ~/.bashrc

# 刷新配置
source ~/.bashrc

Option Three: Windows / macOS Installation

Windows: Directly go to the official website to download the .msi installation package, double-click it to run, and select “Next” until completion. It will automatically configure the environment variables.
macOS: If Homebrew is installed, directly run brew install go.

🚨 Key Step: Configure the Domestic Acceleration Proxy (GOPROXY)

This step is very important!

If not configured, running go install github.com/... in China will almost 100% fail due to a timeout, as the default Go source is blocked.

Please execute the following command in the terminal to enable the domestic proxy (Qiniu Cloud or Alibaba Cloud source):

Bash

# 启用 Go Modules 功能
go env -w GO111MODULE=on

# 配置 GOPROXY 代理 (推荐使用七牛云的 goproxy.cn，速度很稳)
go env -w GOPROXY=https://goproxy.cn,direct

Verify Success

After installation and configuration, run the following command to test:

Check version:

Bash

go version
# 输出示例: go version go1.23.4 linux/amd64

Now you can install s5cmd:

Bash
```
go install github.com/peak/s5cmd/v2@latest
```
(After installation, s5cmd may be located under ~/go/bin/. If PATH is configured earlier, simply typing s5cmd will work)

Notes: Other download methods

This is a standard AWS S3 bucket path. To download the content under this path, traditional “API Key” is usually not required; instead, the AWS CLI command-line tool can be used.

For your three questions, the answers are as follows:

How to download: Use AWS CLI or S3-compatible tools (such as s5cmd).
Is the API required? If it is a public dataset (OpenPI appears to be an open-source project), usually no account or key is needed (use --no-sign-request parameter). If it is private, AWS credentials (Access Key) need to be configured.
Can it be accelerated with multiple threads? Yes. AWS CLI has built-in concurrency capabilities, but using s5cmd tool usually allows full bandwidth utilization and faster speed.

Here are the specific download options, sorted by recommendation level:

Method 1: Using `s5cmd` (Recommended, fastest)

If you are looking for ultra-fast multi-thread download speed, I strongly recommend s5cmd. It is written in Go and has excellent concurrency performance, much faster than the official CLI.

Install s5cmd:

Linux (via go): go install github.com/peak/s5cmd/v2@latest
Linux (direct download): Download the binary file from the release page and extract it.
Conda: conda install -c conda-forge s5cmd

Execute the download (assuming it is a public bucket):

–no-sign-request indicates anonymous access without the need for login.

Bash

s5cmd --no-sign-request cp --concurrency 50 "s3://openpi-assets-simeval/pi0_fast_droid_jointpos/*" ./local_folder/

--concurrency 50: Enable 50 threads for concurrency (you can adjust it according to your bandwidth).

Method 2: Using the official AWS CLI (standard practice)

This is the most general method.

Install AWS CLI:

Bash

pip install awscli
# 或者使用 conda
conda install -c conda-forge awscli

Execute the download:

Since this is a directory, the sync command or cp --recursive should be used.

Case A: If it is a public dataset (Public Bucket)

Typically, the dataset of open-source projects (such as OpenPI) is publicly available without the need for an AWS account login.

Bash

aws s3 sync s3://openpi-assets-simeval/pi0_fast_droid_jointpos ./local_target_folder --no-sign-request

Note: --no-sign-request is crucial; it tells AWS, “I don’t have an account, let me try anonymous downloading.”

Case B: If it is a private dataset

If the above command reports an error “Access Denied”, it means you need permissions. You need to:
1. Register a AWS account.
2. Obtain the Access Key ID and Secret Access Key.
3. Run aws configure and enter the key.
4. Remove the --no-sign-request parameter and run again.

How to speed up AWS CLI:

AWS CLI is multi-threaded by default, but you can optimize performance by modifying the configuration:

Bash
```
aws configure set default.s3.max_concurrent_requests 20
aws configure set default.s3.max_queue_size 10000
```

Method 3: Using Python (`boto3`)

If you want to load directly in the code instead of downloading to the hard drive, you can use the boto3 library in Python.

Python

import boto3
from botocore import UNSIGNED
from botocore.config import Config

# 配置匿名访问 (对应 --no-sign-request)
s3 = boto3.client('s3', config=Config(signature_version=UNSIGNED))

bucket_name = 'openpi-assets-simeval'
prefix = 'pi0_fast_droid_jointpos'

# 列出文件
paginator = s3.get_paginator('list_objects_v2')
pages = paginator.paginate(Bucket=bucket_name, Prefix=prefix)

for page in pages:
    if 'Contents' in page:
        for obj in page['Contents']:
            print(f"Found: {obj['Key']}")
            # 在这里添加下载逻辑: s3.download_file(...)

Summary Recommendations

Preferred: Try aws s3 sync ... --no-sign-request in Method 2. This is the fastest way to verify whether permissions are required.
If the file is large: Be sure to use Method 1 (s5cmd). In cases of large files and many small files, its speed is usually several times that of the official CLI.

Do you need me to write a Python script that automatically checks if the Bucket is public and selects the fastest way to download it?

Option 1: Install using Conda (Recommended; no root privileges required)

Option 2: Standard installation of Linux server (official binary)

Option Three: Windows / macOS Installation

🚨 Key Step: Configure the Domestic Acceleration Proxy (GOPROXY)

Verify Success

Notes: Other download methods

Method 1: Using s5cmd (Recommended, fastest)

Method 2: Using the official AWS CLI (standard practice)

Method 3: Using Python (boto3)

Summary Recommendations

Method 1: Using `s5cmd` (Recommended, fastest)

Method 3: Using Python (`boto3`)