CTRLsum: Towards Generic Controllable Text Summarization by @salesforce in @PyTorch on @Gradio
— AK (@ak92501) May 18, 2021
paper: https://t.co/gIsJNnpB3h
github: https://t.co/s20LBp99jj
gradio demo: https://t.co/PC9fv2j0iI pic.twitter.com/KjIH6nz5XS
This paper gives me anxiety. BatchNorm is the most deviously subtly complex layer in deep learning. Many issues (silently) root cause to it. Yet it is ubiquitous because it works well (it multi-task helps optimization/regularization) and can be fused to affines at inference time. https://t.co/3EC2Abm8Ry
— Andrej Karpathy (@karpathy) May 18, 2021
Rethinking “Batch” in BatchNorm
— AK (@ak92501) May 18, 2021
pdf: https://t.co/ZfLGqlGxPv
abs: https://t.co/oJArBeNN90 pic.twitter.com/TgqI1HkQv7
Divide and Contrast: Self-supervised Learning from Uncurated Data
— AK (@ak92501) May 18, 2021
pdf: https://t.co/Qf5gBiCUP6
abs: https://t.co/Be6bvqiIWM
pretrained on less curated datasets, DnC greatly improves
the performance of self-supervised learning on downstream tasks pic.twitter.com/FhXOy65oMP
Google Replaces BERT Self-Attention with Fourier Transform: 92% Accuracy, 7 Times Faster on GPUs by @Synced_Global https://t.co/wnj2u1QleE
— Bojan Tunguz (@tunguz) May 17, 2021
🦉 I'm really enjoying reading through @DVCorg's documentation and use cases for operationalizing machine learning models.https://t.co/9k8tSfXbMK
— 👩💻 Paige Bailey @ 127.0.0.1 🏡 #BLM (@DynamicWebPaige) May 17, 2021
If you've ever tried to put a model in production, these concepts will resonate. Check out their open-source project on @Github! ✨ pic.twitter.com/zsSdlivwZk
One year of sea surface temperature in the southwestern Indian Ocean. #dataviz
— Randy Olson (@randal_olson) May 17, 2021
Source: https://t.co/8wQjYqXc6J pic.twitter.com/ze4PHeVDrs
"Routes to India" Lovely vintage map from 1893 shows India's international travel connections. High-res version of the map can be found under the link: https://t.co/UApdz918Ah pic.twitter.com/I1ZeXeiVPE
— Simon Kuestenmacher (@simongerman600) May 17, 2021
Project CodeNet: IBM releases 14M sample coding dataset for "AI for code" https://t.co/6G2zcCdj0W
— /MachineLearning (@slashML) May 17, 2021
The Economist built a model to assess the death toll of the pandemic, corrected for underreporting.https://t.co/fJRSRcxtZy
— Max Roser (@MaxCRoser) May 14, 2021
They come to the conclusion that 10 million have died (with an uncertainty of 7 to 13 million).
The map shows the distribution of the death rate. pic.twitter.com/YEwhkuhaEW
Excited to share the open source of EfficientNetV2 code and modes.
— Mingxing Tan (@tanmingxing) May 14, 2021
Paper: https://t.co/YHWEb8pHmR(accepted to ICML'21)
Code: https://t.co/97gKzgYtf2
Bonus: amazing speed on both TPUs and V100/A100 GPUs with TensorRT. https://t.co/rPDTusJWzW
When Does Contrastive Visual Representation Learning Work?
— AK (@ak92501) May 13, 2021
pdf: https://t.co/pw83mXu7Sn
abs: https://t.co/tSHW6h8hVs
benefit of additional pretraining data beyond
500k images is modest, adding pretraining images from
another domain does not lead to more general representations pic.twitter.com/XVHBNvI1bv
