We need more data Can we make it real-time? The business need has… evolved The server is suspiciously slow today Works perfectly on my machine Dataset has 47 samples. Total. Someone hogged all GPUs The labels are completely wrong Who annotated this?! SSH connection died mid- experiment Restart kernel, run all, pray Server crashed during the live demo Stakeholder ghosted after week 2 “Quick 5- min sync” = 45 minutes Pet photobombed the video call Free! A meeting could have been an email CUDA out of memory Actually, let’s pivot the entire approach Forgot to kill my process… sorry team! Jupyter kernel died mysteriously Extreme class imbalance strikes again Test data leaked into training set This should be simple, right? Who left training running ALL weekend? Ground truth isn’t actually true Can we get 99.9% accuracy? Server rebooted overnight. RIP my results. Nice pajama top on the call OOM killed at epoch 99 of 100 Internet died at demo time When will the model be production- ready? This notebook is 2000 cells long I’ll document this later Forgot to shuffle the dataset The data is too noisy to use As per my last email… Requirements changed… again Wait… which cell did I run last? Just add ONE more feature before launch Loss went to NaN on step 1 We need more data Can we make it real-time? The business need has… evolved The server is suspiciously slow today Works perfectly on my machine Dataset has 47 samples. Total. Someone hogged all GPUs The labels are completely wrong Who annotated this?! SSH connection died mid- experiment Restart kernel, run all, pray Server crashed during the live demo Stakeholder ghosted after week 2 “Quick 5- min sync” = 45 minutes Pet photobombed the video call Free! A meeting could have been an email CUDA out of memory Actually, let’s pivot the entire approach Forgot to kill my process… sorry team! Jupyter kernel died mysteriously Extreme class imbalance strikes again Test data leaked into training set This should be simple, right? Who left training running ALL weekend? Ground truth isn’t actually true Can we get 99.9% accuracy? Server rebooted overnight. RIP my results. Nice pajama top on the call OOM killed at epoch 99 of 100 Internet died at demo time When will the model be production- ready? This notebook is 2000 cells long I’ll document this later Forgot to shuffle the dataset The data is too noisy to use As per my last email… Requirements changed… again Wait… which cell did I run last? Just add ONE more feature before launch Loss went to NaN on step 1
(Print) Use this randomly generated list as your call list when playing the game. There is no need to say the BINGO column name. Place some kind of mark (like an X, a checkmark, a dot, tally mark, etc) on each cell as you announce it, to keep track. You can also cut out each item, place them in a bag and pull words from the bag.
We need more data
Can we make it real-time?
The business need has… evolved
The server is suspiciously slow today
Works perfectly on my machine
Dataset has 47 samples. Total.
Someone hogged all GPUs
The labels are completely wrong
Who annotated this?!
SSH connection died mid-experiment
Restart kernel, run all, pray
Server crashed during the live demo
Stakeholder ghosted after week 2
“Quick 5-min sync” = 45 minutes
Pet photobombed the video call
Free!
A meeting could have been an email
CUDA out of memory
Actually, let’s pivot the entire approach
Forgot to kill my process… sorry team!
Jupyter kernel died mysteriously
Extreme class imbalance strikes again
Test data leaked into training set
This should be simple, right?
Who left training running ALL weekend?
Ground truth isn’t actually true
Can we get 99.9% accuracy?
Server rebooted overnight. RIP my results.
Nice pajama top on the call
OOM killed at epoch 99 of 100
Internet died at demo time
When will the model be production-ready?
This notebook is 2000 cells long
I’ll document this later
Forgot to shuffle the dataset
The data is too noisy to use
As per my last email…
Requirements changed… again
Wait… which cell did I run last?
Just add ONE more feature before launch
Loss went to NaN on step 1