"Paraphrastic representations at scale" is a strong, blazing fast package for sentence embeddings by @johnwieting2
— Graham Neubig (@gneubig) June 22, 2021
Paper: https://t.co/LlHbH788b2
Code: https://t.co/91XR75PDG9
Beats Sentence-BERT, LASER, USE on STS tasks, works multilingually, and is up to 6,000 times faster 😯 pic.twitter.com/julWhnjyYf
HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units by @facebookai in @PyTorch on @Gradio Hub using @huggingface models
— AK (@ak92501) June 16, 2021
paper: https://t.co/OzIO3tL351
github: https://t.co/g691cqN7Fb
gradio demo: https://t.co/46qkxR59n4 pic.twitter.com/G79R9z2X6X
We just released the multilingual C4 dataset! 1.4TB of German? 1.6TB of Spanish? 3.6TB of Russian 😲? We got it all. Check out the announcement at https://t.co/T6wcPDl5O7!
— AllenNLP (@ai2_allennlp) June 16, 2021
🚀 And merged to Transformers!
— Hugging Face (@huggingface) June 1, 2021
We are excited to welcome ByT5 as the first tokenizer-free model!
👉All available checkpoints can be accessed on the 🤗hub here: https://t.co/WEpfzu6uMN
👇 Demo (on master): https://t.co/2qgmt7YQO8 pic.twitter.com/m5pXe9qETU
CogView: Mastering Text-to-Image Generation via Transformers
— AK (@ak92501) May 28, 2021
pdf: https://t.co/SOIqFhHleW
abs: https://t.co/nmhinD6kXK
demo: https://t.co/QULIeN628P
a 4-billion-parameter Transformer with VQ-VAE tokenizer, achieves a new SOTA FID on blurred MS COCO pic.twitter.com/hvXheo5cYP
Introducing a simple one-stage approach for automated text-to-image generation that achieves state-of-the-art performance on #LocalizedNarratives using only image-text pairs — no segmentation or bounding box data needed. Learn how it’s done below: https://t.co/SPMbJaVvAN
— Google AI (@GoogleAI) May 26, 2021
True Few-Shot Learning with Language Models
— AK (@ak92501) May 25, 2021
pdf: https://t.co/a0WvisxlJX
abs: https://t.co/7zb714iSYk
github: https://t.co/nPICjpoRuo
few-shot ability of LMs when held-out examples are unavailable, prior work significantly overestimated the true few-shot ability of LMs pic.twitter.com/k28CCi4iWk
Today we are announcing our work on building speech recognition models without any labeled data! wav2vec-U rivals some of the best supervised systems from only two years ago.
— Michael Auli (@MichaelAuli) May 21, 2021
Paper: https://t.co/cYzF9MGu56
Blog: https://t.co/iiGmgdnCiV
Code: https://t.co/TQ56tT0unx
Check out KELM, a method of converting knowledge graphs to synthetic natural language sentences, leveraging their factual nature in order to augment existing #NLP pre-training corpora. Learn more and access the corpora at ↓ https://t.co/jbrDVHCgeF
— Google AI (@GoogleAI) May 20, 2021
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
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
Are Pre-trained Convolutions Better than Pre-trained Transformers?
— AK (@ak92501) May 10, 2021
pdf: https://t.co/8L06XiPM1C
abs: https://t.co/gIAq2Od5GA
experimental results show that convolutions can outperform Transformers in both pretrain and non-pre-trained setups pic.twitter.com/IZtYlBXvkc
