How to fix the error "FATAL: container creation failed:" "error: can't mount image " "failed to mount squashfs filesystem: input/output error":
#This occurs when you try to run a singularity container within a VM and it is located on a mounted ntfs filesystem. The solution is to transfer the container to a different filesystem eg. ext3.
How to negotiate a better home loan rate with your bank (from Barefoot Investor SCOTT PAPE):
You: Hello, my account number is ______. I’ve been with you for ___ years, but I’ve applied to refinance with UBank. Their rate is ____ per cent, which is a full ___ per cent cheaper than you’re charging me. Given our longstanding relationship, I’d like you to match the offer — or send me the forms I need to switch to UBank.
Bank rep: One moment, please.
(You’re bluffing, of course. However, the bank’s sales team have strict targets, backed by incentives, that they have to meet — one of which is giving profitable customers discounts to stop them leaving.)
Bank rep: We can’t match the rate you have quoted. However, we understand you are a valuable customer, so we would like to offer you a 0.15 per cent discount.
You: That’s not good enough. I’ve already got conditional approval … so in order to stay I need at least a 0.5 per cent discount. Could you please speak to your supervisor? I’m happy to wait.
Bank rep (a full six minutes later): On reviewing your case, we can offer you that 0.5 per cent discount on your current rate.
You: Brilliant! Please send me an email confirming the new rate and confirming that it will be applied as of start of business tomorrow.
How to visualise massive point clouds in python:
https://towardsdatascience.com/guide-to-real-time-visualisation-of-massive-3d-point-clouds-in-python-ea6f00241ee0
How to install point-e on Windows 11:
#in Windows Powershell note down the cuda version from this command:
nvidia-smi.exe
#look up the channel for your cuda version using this site:
#https://anaconda.org/nvidia/repo/installers?
#In Anaconda powershell:
conda update conda
conda update anaconda
conda update python
conda update --all
conda create --name point-e pip
#Now exit from the Powershell you are using and then open a new one before you activate the new "env".
conda activate point-e
#install cuda from the channel identified previously
conda install cuda --channel "nvidia/label/cuda-12.0"
#NB. there doesn't seem to be an option (yet) to use cud=12.0 in the code line below. The line below is suggested when visiting https://pytorch.org/
conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c nvidia
#NB. I found that the line below was necessary to get things to work
conda install cudatoolkit
conda install ipykernel jupyter
nvcc --version
conda install -c anaconda git
git clone https://github.com/openai/point-e.git
cd point-e
pip install -e .
jupyter notebook
#navigate to the notebook in Jupyter at:
#./point-e/point_e/examples/text2pointcloud.ipynb
#In the notebook add a cell at the beginning with the code below to test if the GPU is detectable:
import torch
print(f'PyTorch version: {torch.__version__}')
print('*'*10)
print(f'_CUDA version: ')
!nvcc --version
print('*'*10)
print(f'CUDNN version: {torch.backends.cudnn.version()}')
print(f'Available GPU devices: {torch.cuda.device_count()}')
print(f'Device Name: {torch.cuda.get_device_name()}')
#NB. I had an issue with the kernel crashing when plotting the pointclouds.
#The error in the anaconda powershell was:
#OMP: Error #15: Initializing libiomp5md.dll, but found libiomp5md.dll already initialized.
#The solution was to add a cell at the start of the script with the code:
import os
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
#NB. The above workaround apparently may cause crashes or silently produce incorrect results....
