Home
Popular
I'm having trouble running this line of code, does anyone know why?Β
np.random.seed(41)
np.random.seed(41)
data = ImageDataBunch.from_folder(path, train="train", valid ="test",ds_tfms=get_transforms(), size=(256,256), bs=32, num_workers=4).normalize()
The error says:
ο»Ώ
And previously, I have imported the Fastai library.
The error says:
ο»Ώ
And previously, I have imported the Fastai library.
Google drive integration would be sweet as we can upload our notebooks to cloud also more extensions and themes would be welcomed
EDIT : Would love to see csv files in tabular format while working with it and to open tabs inside the deepnote
EDIT : Would love to see csv files in tabular format while working with it and to open tabs inside the deepnote
I would like to share notebooks as reports for the layman without any code, but just the graphics and tables and markdown.
I hid the cell output, and then published my notebook. I found that the published notebook had tiny lines for every cell I had hidden code for (for the cells that had no output). See image below.
I like the publishing feature, but I want it without any code as I want to author reports that are not targeted at people who care about the code.
Also, it would be nice if there was a feature that could easily hide/show all code for all cells in a notebook, and not one by one.
Please share your thoughts - maybe I didn't do something right and this is all possible?ο»Ώ
I hid the cell output, and then published my notebook. I found that the published notebook had tiny lines for every cell I had hidden code for (for the cells that had no output). See image below.
I like the publishing feature, but I want it without any code as I want to author reports that are not targeted at people who care about the code.
Also, it would be nice if there was a feature that could easily hide/show all code for all cells in a notebook, and not one by one.
Please share your thoughts - maybe I didn't do something right and this is all possible?ο»Ώ
You can try to train a huge machine learning thing to estimate execution time based on cell code + memory consumption by variables used there + characteristics of the machine.
Ways to train that:
Ways to train that:
- Publicly available GitHub notebooks, which you can execute and collect all required data like variables size. This variant gets a mark A+ from me: it does not use private data -> no legal issues, but you'll need to execute all these notebooks. Considering they're not "reactive", about 50% of notebooks will be lost (but there is nothing to do with that)
- Actually, you have an enormous number of statistics on the usage of code from Deepnote itself, so that might work. I don't know if that fits your terms of usage, but if no, then you can add a special toggle in settings like "use my anonymized data to enhance UX". But don't know if that is ok, but you actually won't need to execute notebooks, as they're executed by the user himself.
Maybe you can split that into several parts and code on Twitch stream. For example:
- Collecting all notebooks from GitHub
- Writing a thing to execute them and collect data
- ...
As that will be a hard task to estimate precise time, then you can make 3 (or more) categories:Β
- Will finish in less than 1 minute
- Will finish in 1-5 minutes
- Will finish in 5+ minutes
That has its benefits:
- Β You won't need to wait for a cell with "while (True)" to finish, you can stop after 5 minutes of execution
- Precise time might be not so important for me as a user: I just want to understand if I should sit here for 3 mins and wait or I can go get some coffee
As per previous posts, we have multiple notebooks per project. Often, what is different from one notebook to another are experiment settings (hyperparameters etc), but they all share in common the same core code base which are simply imported Python files.
It'd be nice if we could run all our notebooks after making changes to the common code base, perhaps a "Run all" or allow us to pick and choose which ones. Thank you!
It'd be nice if we could run all our notebooks after making changes to the common code base, perhaps a "Run all" or allow us to pick and choose which ones. Thank you!
As the title says, how long will it take before ipywidgets support is added to Deepnote? It has been a pain point for a while, and has forced me to download notebooks and run them locally in more than a few instances.
Hi, can someone provide me with a pro/con list or parameter of Google Colab vs. Deepnote? My list so far:
DeepNote
General
- customer service
Resources
- 5gb ram free compute
UI and features
- beautiful design
- click-optimized UI (powerful hardware with one click)
- real time collaboration
- environment management
- quick integration with AWS, Mongo, etc.
- SQL cells
- reactive apps with forms
- automatic visualizationΒ
Google Colab
General
- big community/network
Resources
- 12gb ram free compute
- 11gb free GPU memory
- TPU (180 TFlops, 64gb HBM)Β
- notebooks can run for at most 12 hours
UI and features
- FormsΒ
- code snippets
Happy to clarify those points and complete the list together!
CC ο»Ώ
%20cropped.jpg)
I want to install quanteda once for a project and have it be available for all the notebooks in that project. To do this I modify the docker file of the R 3.5.2 template as follows:
FROM gcr.io/deepnote-200602/templates/deepnote
RUN sudo apt-get update && \ sudo apt-get install -y r-base libxml2-dev libcurl4-openssl-dev libssl-dev libmariadbclient-dev libgit2-dev
RUN sudo R -e "install.packages(c('IRkernel', 'quanteda'), repos='http://cran.rstudio.com/' , dependencies=TRUE)"
RUN R -e "IRkernel::installspec()"
RUN sudo usermod -a -G staff jovyan
# Set default kernel to R
ENV DEFAULT_KERNEL_NAME "ir"
Here is my notebook:Β
My problem is that the build times out. Any recommendations? Thank you for your help.
Having problems getting the GCS integration to work. GQB integration works fine, using service accounts.
After setting up, /datasets/my-bucket-name gets mounted, but `!ls` doesn't show any files nor directories.
After setting up, /datasets/my-bucket-name gets mounted, but `!ls` doesn't show any files nor directories.
This month's community update is special - as you know, the first Deepnote publishing competition is over! π
We've had over 40 submissions from more than 30 of you masterminds. Thanks to every single one of you for all your hard work & sharing it with the community! Our panel read each notebook thoroughly and evaluated all submissions using 4 criteria:
We've had over 40 submissions from more than 30 of you masterminds. Thanks to every single one of you for all your hard work & sharing it with the community! Our panel read each notebook thoroughly and evaluated all submissions using 4 criteria:
- Explainability
- Creativity of the research question
- Use of tools
- Presentation
We were absolutely blown away by the quality and creativity of your work, so instead of announcing 4 winners, we've decided to award top 8 projects, and added extra prizes. Without further ado, here are the winners! π
π The winner
1οΈβ£ The $1,000 grand prize goes to ο»Ώ
Marc Bernot
ο»Ώ who blew our minds by reverse-engineering the MoirΓ© effect on San Pellegrino water bottle labels using the xarray and holoviews libraries. Please give it up for Mark π and check out his notebook here: https://deepnote.com/@marc-bernot/San-Pellegrino-label-moire-effect-sWIn2nYPSeuvU7na8_SxBQ
β¨ Finalists
All the finalists in this section will receive 1-year subscription to the Deepnote Teams plan, a swag pack and a $100 Amazon gift card. Congrats!!
2οΈβ£ On #2, we have ο»Ώ
Tomas Bonfiglio
ο»Ώ who showed us how to enhance clustering projects using Scikit Learn, leveraged some advanced Deepnote feature and deployed the result as a Streamlit app: https://deepnote.com/@tomas-bonfiglio/Enhance-your-clustering-projects-T176oVvEQZWXnoZZkusECA
3οΈβ£ The top 3 is topped off by ο»Ώ
Ingrid Chien
ο»Ώ who created a beautiful showcase comparing the harmonious data science lifecycle with that portrayed in Avatar The Last Airbender ππ₯π¬οΈπ : https://deepnote.com/@ingrid-chien-5a46/Avatar-The-Last-Airbender-Connecting-Data-Science-to-the-Four-Forces-of-Nature-n02JIZmRQyWbW7tg_Znlpw
4οΈβ£ - ο»Ώ
Franccesco Orozco
ο»Ώ showed us how to build a customer segmentation and cohort analysis, explaining in great detail how you can replicate it yourself to understand customer behavior and drive strategic insights at your organization: https://deepnote.com/@Radial-Inc/Customer-Segmentation-and-Behaviors-nbIw_gpUT_uLZTcSQuV5QA
β Semifinalists
All semifinalists will receive 1-year subscription to the Deepnote Teams plan, a swag pack and a $50 Amazon gift card. Well done!
5οΈβ£ - "Would it be possible to generate those pencil-like drawings with random forests, with different types of trees?" ο»Ώ
Daniel Kapitan
ο»Ώ asked himself and went on to create beautiful generative art with L-systems in Python: https://deepnote.com/@dkapitan/lartpy-YJpZGp2rQE66bpM5-0-80Q
6οΈβ£ - For all reinforcement learning fans, ο»Ώ
Martin Ferianc
ο»Ώ & Xiaoliang Luo built a comprehensive intro to RL using the OpenAI Gym library and solved a concrete problem using the classic Q-learning algorithm: https://deepnote.com/@ken-e7bd/Intro-to-Q-learning-in-RL-4R450s6_RVKJIC2xiqs71g
7οΈβ£ - ο»Ώ
Chanakya Vivek Kapoor
ο»Ώ shows how to generate synthetic data using Gaussian Mixture Model: https://deepnote.com/@chanakya-vivek-kapoor/Synthetic-Data-Generation-QaaTRs73T2iCb0amHFbwpQ
8οΈβ£ - ο»Ώ
Abid
ο»Ώ used Tripadvisor data describing customers' hotel experience and built a model predicting ratings based on reviews. He also used NLP tools for feature extraction and preparing the data for deep learning: https://deepnote.com/@abid/Trip-Advisor-Data-AnalysisML-9gYLOddsRXmWSKVLyLX_tQ
π Competition MVP
π Competition MVP
A special shoutout goes to ο»Ώ
Abid
ο»Ώ, who submitted over 10 notebooks for this round of competition! We're sending 30 hours of Pro machines your way so that you make more awesomeness happen.
π π Congratulations again to our winners and thank you all again for your hard work, we had so much fun reading through your projects. We'll be back with another version of the competition soon. In the meantime, we'd love to invite everyone to another Deepnote community Happy hour next week to celebrate and congratulate you all on your great work! You can RSVP here.Β
