TIL you can reference a Python variable in a bash cell of a jupyter notebook and use that to do some really crazy mixed mode scripting. pic.twitter.com/YgAq43v0GB
— Delip Rao (@deliprao) August 26, 2020
TIL you can reference a Python variable in a bash cell of a jupyter notebook and use that to do some really crazy mixed mode scripting. pic.twitter.com/YgAq43v0GB
— Delip Rao (@deliprao) August 26, 2020
What tools should an R data scientist master to write performant code? My top three:
— David Robinson (@drob) August 26, 2020
1. dplyr/data.table
2. Vector/matrix operations
3. dbplyr (or another DB ORM)
First 2 get you 80/20 (respectively) for fast data transformations, and 3rd gets you scale https://t.co/Levlv8NdeQ
Communication tip: When you're writing for an audience of varying experience levels, explain concepts using language that to experts doesn't feel like an explanation pic.twitter.com/llCkTr4GWy
— David Robinson (@drob) August 24, 2020
Tip: when running on TPU, you can significantly speed up your model by running *multiple steps of gradient descent in a single graph execution*. This helps get the device to 100% utilization (which, for a TPU, is huge).
— François Chollet (@fchollet) August 18, 2020
Just pass `experimental_steps_per_execution` to `compile`. pic.twitter.com/cwzk27z5Lo
Today’s pro leadership tip:
— Angela Bassa (@AngeBassa) August 18, 2020
You don’t have to work with assholes. You can choose to surround yourself with kind people 🥰
Recommend this 🔑 article on being a 10x data scientist.
— Lavanya (@lavanyaai) August 18, 2020
Includes 💯 nuggets like:
"The only way to build trust with the scientific community is to commit to radical transparency ... post all your training runs to a public dashboard via @weights_biases."https://t.co/Os9GYQ9J5h
I say this in talks all the time, don’t be afraid to ask users vs making a model. Usually this happens because the ML team has been pushed down in the org or act in an embedded service model such that they have no real control over UX. Front end changes for ML become impossible.
— peteskomoroch (@peteskomoroch) August 16, 2020
.@thomas_mock's taking parameterized reporting with RMarkdown to the next level - use purrr::walk to render multiple reports, each named correctly, for a list of parameters #rstatsnyc pic.twitter.com/9wUFrnXbTG
— Emily Robinson (@robinson_es) August 15, 2020
🤯 Did you know you could reference blocks of code in an .R file from RMD by referencing them by name??@thomas_mock talk (slides at https://t.co/fDBJPbwTGC) is full of great RMarkdown tips - check it out! #rstatsnyc pic.twitter.com/edFo6pdac2
— Emily Robinson (@robinson_es) August 15, 2020
Did you know you can append data to an existing CSV file without opening that file and importing the data into #Rstats first? Use append = TRUE in either #rdatatable's fread() function or the #tidyverse vroom pkg's vroom_write(). pic.twitter.com/Md3RYUiplh
— Sharon Machlis (@sharon000) August 13, 2020
Q: How do I get a top job/PhD in Deep Learning?
— Andrew Trask (@iamtrask) August 10, 2020
A: Get experience! Demonstrate your ability!
Q: How?
A: Join an open-source project, especially in a new/growing research area!
Q: How?
A: Here's 50+ opportunities... grouped by typehttps://t.co/1pIGQdkplN #100DaysOfMLCode pic.twitter.com/KBAERXRY3C
I've heard a lot of really successful https://t.co/GEOZuodrZj alums say this is the most important advice they got from my teaching.
— Jeremy Howard (@jeremyphoward) August 10, 2020
Pick a project. Make it awesome. Polish it and love it. https://t.co/1fmpJI3YcT