Shiwen An

Shiwen An

Someone working hard on Quantum Computing and AI security. Occasionally Blogger, Kaggler, and AtCoder.

Posts by Tags

2dmatrix

AI Infrastructure

Fine-Tuning LLaMA 2 with torchrun

1 minute read

Published:

Fine-tuning large models like LLaMA 2 is a big task, but with torchrun, you can scale it across multiple GPUs with ease. In this post, I’ll walk you through how to do it step by step.

Algorithm

Tensor Train Decomposition and Training

11 minute read

Published:

High-dimensional data (tensors) appear in many fields such as scientific computing, quantum physics, and machine learning. However, storing and operating on these tensors is challenging due to the exponential growth of parameters with the number of dimensions (the so-called “curse of dimensionality”). Tensor Train (TT) decomposition is one way to represent high-dimensional tensors in a compact format by expressing them as a sequence of smaller 3D tensors (often called TT-cores).

Fine-tuning

Fine-Tuning LLaMA 2 with torchrun

1 minute read

Published:

Fine-tuning large models like LLaMA 2 is a big task, but with torchrun, you can scale it across multiple GPUs with ease. In this post, I’ll walk you through how to do it step by step.

Machine Learning

Tensor Train Decomposition and Training

11 minute read

Published:

High-dimensional data (tensors) appear in many fields such as scientific computing, quantum physics, and machine learning. However, storing and operating on these tensors is challenging due to the exponential growth of parameters with the number of dimensions (the so-called “curse of dimensionality”). Tensor Train (TT) decomposition is one way to represent high-dimensional tensors in a compact format by expressing them as a sequence of smaller 3D tensors (often called TT-cores).

Parallel Computing

Fine-Tuning LLaMA 2 with torchrun

1 minute read

Published:

Fine-tuning large models like LLaMA 2 is a big task, but with torchrun, you can scale it across multiple GPUs with ease. In this post, I’ll walk you through how to do it step by step.

Tensor

Tensor Train Decomposition and Training

11 minute read

Published:

High-dimensional data (tensors) appear in many fields such as scientific computing, quantum physics, and machine learning. However, storing and operating on these tensors is challenging due to the exponential growth of parameters with the number of dimensions (the so-called “curse of dimensionality”). Tensor Train (TT) decomposition is one way to represent high-dimensional tensors in a compact format by expressing them as a sequence of smaller 3D tensors (often called TT-cores).

beginner

  1. Servo Book Documentation Update on CanGc

less than 1 minute read

Published:

The Update on the CanGc in the documentation for developers. Some time the borrow hazard could occur when assigning the mutable variable. Looks like this is a long term issue happened repeatedly when CanGc need to pass in as arguments. Several issues here listed.

  1. First Issue from the Servo Team (Update)

1 minute read

Published:

This time is the record for updating again the first issue on line number, because the team official think my solution is not good enough and provide other solution here.

cache

conda

Install Miniconda on Server

less than 1 minute read

Published:

Step 1: Download Miniconda Locally

Run the following command to download the latest Miniconda installer for Linux (adjust the link if using macOS):

Python Env for Scientific Experiment

3 minute read

Published:

As we usually need to work on various project, the environment for the experiment some times different from the local environment. Docker is one of the option, to run on the virtual machine. However, Conda or mamba is more like the standard for current academia.

container

Using Docker within the VS Code

1 minute read

Published:

When trying to include everything, including the environment for development might not be enough, even with the same ubuntu system. Recently, I encountered a bug where display is not properly shown, I have several doubts regarding the scenario.

docker

Docker Container Debug with VScode

2 minute read

Published:

When using the docker container, one may want to use the feature in the vscode to remotely work on the code debug. For personal server, this would guarantee a relatively clean environment together with the debug feature.

Utilize the Docker container

4 minute read

Published:

CUDA Docker Container Setup and Usage Guide This tutorial covers how to build, run, attach, and detach a CUDA-enabled Docker container supporting three NVIDIA A6000 GPUs.

dp

env

Docker Container Debug with VScode

2 minute read

Published:

When using the docker container, one may want to use the feature in the vscode to remotely work on the code debug. For personal server, this would guarantee a relatively clean environment together with the debug feature.

Utilize the Docker container

4 minute read

Published:

CUDA Docker Container Setup and Usage Guide This tutorial covers how to build, run, attach, and detach a CUDA-enabled Docker container supporting three NVIDIA A6000 GPUs.

Install Miniconda on Server

less than 1 minute read

Published:

Step 1: Download Miniconda Locally

Run the following command to download the latest Miniconda installer for Linux (adjust the link if using macOS):

Python Env for Scientific Experiment

3 minute read

Published:

As we usually need to work on various project, the environment for the experiment some times different from the local environment. Docker is one of the option, to run on the virtual machine. However, Conda or mamba is more like the standard for current academia.

fsm

grid

intro

jit

Using Docker within the VS Code

1 minute read

Published:

When trying to include everything, including the environment for development might not be enough, even with the same ubuntu system. Recently, I encountered a bug where display is not properly shown, I have several doubts regarding the scenario.

llvm

LLVM learning note by hands

less than 1 minute read

Published:

Start with tutorial from the official website should be the only route to expertise, things changed up and down.

memory

optimization

LLVM learning note by hands

less than 1 minute read

Published:

Start with tutorial from the official website should be the only route to expertise, things changed up and down.

performance

LLVM learning note by hands

less than 1 minute read

Published:

Start with tutorial from the official website should be the only route to expertise, things changed up and down.

pyTorch

Fine-Tuning LLaMA 2 with torchrun

1 minute read

Published:

Fine-tuning large models like LLaMA 2 is a big task, but with torchrun, you can scale it across multiple GPUs with ease. In this post, I’ll walk you through how to do it step by step.

python

Tensor Train Decomposition and Training

11 minute read

Published:

High-dimensional data (tensors) appear in many fields such as scientific computing, quantum physics, and machine learning. However, storing and operating on these tensors is challenging due to the exponential growth of parameters with the number of dimensions (the so-called “curse of dimensionality”). Tensor Train (TT) decomposition is one way to represent high-dimensional tensors in a compact format by expressing them as a sequence of smaller 3D tensors (often called TT-cores).

Docker Container Debug with VScode

2 minute read

Published:

When using the docker container, one may want to use the feature in the vscode to remotely work on the code debug. For personal server, this would guarantee a relatively clean environment together with the debug feature.

Utilize the Docker container

4 minute read

Published:

CUDA Docker Container Setup and Usage Guide This tutorial covers how to build, run, attach, and detach a CUDA-enabled Docker container supporting three NVIDIA A6000 GPUs.

Install Miniconda on Server

less than 1 minute read

Published:

Step 1: Download Miniconda Locally

Run the following command to download the latest Miniconda installer for Linux (adjust the link if using macOS):

Python Env for Scientific Experiment

3 minute read

Published:

As we usually need to work on various project, the environment for the experiment some times different from the local environment. Docker is one of the option, to run on the virtual machine. However, Conda or mamba is more like the standard for current academia.

recursion

rust

  1. Servo Book Documentation Update on CanGc

less than 1 minute read

Published:

The Update on the CanGc in the documentation for developers. Some time the borrow hazard could occur when assigning the mutable variable. Looks like this is a long term issue happened repeatedly when CanGc need to pass in as arguments. Several issues here listed.

  1. First Issue from the Servo Team (Update)

1 minute read

Published:

This time is the record for updating again the first issue on line number, because the team official think my solution is not good enough and provide other solution here.

state

subtree

transitions

tree

vscode

Docker Container Debug with VScode

2 minute read

Published:

When using the docker container, one may want to use the feature in the vscode to remotely work on the code debug. For personal server, this would guarantee a relatively clean environment together with the debug feature.

xor