How Many Discoveries Were Lost Because They Weren't Written Down?

With Reports you can save your ingenious ideas and insights alongside a snapshot of your work and then share them with your team. Or just keep them all for when you need them.

Scott Condron

Created on November 25|Last edited on April 3

Comment

﻿
👋📎  Hey, I'm a meta-Report, a Report about Reports!
💡
Reports are for your brain dumps 🧠, eureka moments 💡and everything in between. They play nicely with everything else in W&B so you can share, store or explain your findings. 
📊 Share your experiments chartsImagine arriving at a colleague's folder of random charts and spreadsheets and being asked to make sense of them. Be the change you want to see in the world, Reports exist so you never inflict that pain on yourself or anyone (again 🤐)!
Before W&B Reports:
Ok, this is a bit extreme... but you get the point 😁
The charts below are live and interactive. They are fed with the metrics and configs coming from your W&B runs.
Below the plot, you can click on the Run set to navigate to the run that produced each line in these charts. 
﻿
Run set101
﻿
🧐 Show different views and snapshots of your runs 🤳The chart below is filtered to remove null losses and sorted by loss. When you make one of these beautiful charts, jot down your thoughts so you know what you were trying to investigate and explain to the reader why? 
This allows the reader to digest that insight more easily and invites them to ask questions, but also allows them to pick up where you left off. By using Reports, you can decouple who runs the experiment from who does the analysis.
Comment on PlotsYou can write comments by clicking the top right of a plot (either the speech bubble 💬 or the yellow circle 🟡).
🥶 Freeze your runs to stop any new runs from being added to your plot by clicking the snowflake. ❄️
💡
﻿
﻿
Run set12
﻿
👟 Share your model predictions with W&B TablesStore your predictions at a point in time. Make these Tables yourself with W&B Tables. Don't forget you can filter, group and sort these tables too!
﻿
Run set101
﻿
💄 Why not make your discoveries shine with nice formatting?Alongside any of your W&B panels, you can also use lots of fancy "nice to have" features to improve your Reports for your teammates.
📋 Quickly format your thoughts with Markdown Dear Diary,
Today I found a bug in one of my research papers from 15 years ago that no one ever noticed.
﻿
😎 Code, of course.for xb, yb in dataloader:
    yb_pred = model(xb)
    loss = criterion(y_pred, y)
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()
👩‍🏫 Explain the details with LaTeX﻿f(x)={1if  w⋅x+b>0,0otherwise\displaystyle f(\mathbf {x} )={\begin{cases}1&{\text{if }}\ \mathbf {w} \cdot \mathbf {x} +b>0,\\0&{\text{otherwise}}\end{cases}}f(x)={10​if  w⋅x+b>0,otherwise​﻿﻿
☑️ Checklists because who doesn't love TODOsTODO:
Make a todo list
Check off a todo
Solve the AI control problem 💀
📸 Images and Gifs but don't dare include a matplotlib loss plot or we'll ban you!pronounced JIF 
Party parrot because we're "with it"
﻿
﻿
﻿
﻿
Learn more:Watch this video or read our docs: https://docs.wandb.ai/guides/reports﻿
﻿
﻿
📚 More lovely reports
DALL-E Mini Explained
Generate images from a text prompt in this interactive report: DALL-E Mini Explained, a reproduction of OpenAI DALL·E
An Introduction to Weights & Biases Artifacts
This article introduces Weights & Biases Artifacts, which offer lightweight dataset versioning, and gives an easy-to-follow example so that you can see how they work. 
Omnimatte: How to Detect Objects and Their Effects
Automatically capture the shadows and reflections of objects in videos
Tables Tutorial: Visualize Data for Image Classification
How to version and interactively explore data and predictions across train/val/test with W&B's new Tables feature
﻿
Visit our blog!Check out our blog, Fully Connected, to keep up to date with Machine Learning and W&B: wandb.ai/fc﻿
﻿