Yujin Tang
@yujin_tang
Followers
1,583
Following
764
Media
23
Statuses
590
Explore trending content on Musk Viewer
音楽の日
• 352752 Tweets
Carre
• 262065 Tweets
ダンスバトル
• 246007 Tweets
MUSE HIGHLIGHT MEDLEY
• 163693 Tweets
#LaVeladaDelAñoIV
• 159510 Tweets
#DursunÖzbekTransferlerNerde
• 135450 Tweets
FAYEYOKO WATCH BLANK
• 114495 Tweets
La Cobra
• 103251 Tweets
#1YearWithSeven
• 92823 Tweets
Alana
• 91334 Tweets
GIANT POP STAR JUNGKOOK
• 88927 Tweets
Be Mine
• 79616 Tweets
#SixTONESANN
• 78505 Tweets
Bernabéu
• 71655 Tweets
Guanyar
• 52283 Tweets
Paulo Londra
• 46388 Tweets
Ibai
• 43728 Tweets
Agustín
• 41983 Tweets
Rivers
• 33985 Tweets
Onana
• 33797 Tweets
ATRI
• 28618 Tweets
Julieta Venegas
• 26852 Tweets
Young Miko
• 25979 Tweets
ROCKSTAR TAERAE DAY
• 18155 Tweets
Dr. Ruth
• 16576 Tweets
Erden Timur
• 15919 Tweets
Pogacar
• 15866 Tweets
İzmir
• 14390 Tweets
OUR FIRST SUMMER WITH JIYOON
• 12642 Tweets
#RSAvIRE
• 12318 Tweets
Luis Miguel
• 11548 Tweets
Pinned Tweet
Successfully completed my pet project (pun intended)! The 🤖🐕 can follow commands, navigate to target positions, kick a soccer ball, and play Taiko-no-Tatsujin 🥁
Next project: school of fish 🐟🐠🐡
4
19
161
So, some of the humanoid agents we trained in the past actually demonstrated "normal gaits"
1
10
172
New paper w/
@hardmaru
on applying attention for RL after AttentionAgent (), this time we shift our attention to the sensory neuron level. The agent's not only permutation invariant, but also robust to noises and generalizes better.
The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning
We explore RL agents that still work even when their observations get shuffled around a lot!
A fun paper w/
@yujin_tang
web
pdf
28
240
1K
2
23
157
pfrl from
@PreferredNetJP
is a great library for RL training that is based on PyTorch. I used it to reproduce the results from OpenAI’s Procgen paper:
1
26
150
In this task, all fishes move at constant speed, they observe the states of the nearest M fishes and learn to orient so that they move in a coherent fashion. All fishes share the same MLP policy.
EvoJAX allows quick integration of new tasks, we hope to see more user extensions.
My intern efforts have been merged🎉
Based on the flocking example from jax-md <>, we have migrated the flocking movement to work in the EvoJAX environment.
check it out!
3
9
95
1
25
123
We release a new demo task in EvoJAX:
A group of agents (blue) learnt to hunt another (red). An agent gets air from N,S,E,W neighbors if they are not occupied by an opponent or a wall, and dies if its total air is less than 2.
5
24
116
SayTap: Language to Quadrupedal Locomotion
We use foot contact patterns as interface to bridge instructions in NL and low-level control commands.
New paper w/ Wenhao Yu, Jie Tan,
@heiga_zen
,
@AleksandraFaust
,
@ttyharada
Web
PDF
3
25
92
Our paper is accepted for a spotlight presentation at NeurIPS 2021!
New paper w/
@hardmaru
on applying attention for RL after AttentionAgent (), this time we shift our attention to the sensory neuron level. The agent's not only permutation invariant, but also robust to noises and generalizes better.
2
23
157
6
6
82
We have well-developed infra and toolkits for RL, but the equivalent is missing for neuroevolution practitioners. We hope EvoJAX can fill this gap. Your feedback and contributions are most welcome!
EvoJAX is developed by
@yujin_tang
@alanyttian
We tried to make evolution run really fast with JAX on a wide range of tasks: MNIST, Seq2Seq, Locomotion, Multi-Agent Competition, Generative Art.
WaterWorld-Env adapted from
@karpathy
’s old JavaScript demo!
4
17
93
1
14
62
Our paper “DEIR: Efficient and Robust Exploration through Discriminative-Model-Based Episodic Intrinsic Rewards'' has been accepted at IJCAI 2023 (15% acceptance rate, w/
@swan_104
@tkaneko
@alanyttian
). DEIR explores more efficiently, especially in partially observable tasks.
6
13
61
Fusing Gundam with Transformer is the "Gotenks" magic in the model space!
By merging a JP LM with specialist models, we achieve top perf on various JP LM benchmarks.
We explored param and data flow space merging, and will further propel the tech across modalities and functions.
Introducing Evolutionary Model Merge: A new approach bringing us closer to automating foundation model development. We use evolution to find great ways of combining open-source models, building new powerful foundation models with user-specified abilities!
55
420
2K
0
13
56
Discovering EvoLLM: Harnessing LLMs as evolutionary operators unveils remarkable insights. Thanks
@RobertTLange
for pioneering this work.
Merging evolutionary algorithms & LLMs could unlock a realm of exciting opportunities. A promising avenue for future exploration! ✨
🎉 Happy to share my internship project
@GoogleDeepMind
🗼 – purely text-trained LLMs can act as evolutionary recombination operators 🦎
🧬 Our EvoLLM uses LLM backends to outperform competitive baselines. Work done w.
@alanyttian
&
@yujin_tang
🤗
📜:
5
24
166
0
5
40
Inspired by
@Troika_London
's work, I used EvoJAX () to create a visual illusion in sim, where a chain of particles place themselves in space to show a square from the front and a heart when observed from the back. The video below shows an initial result.
2
3
32
Our work has been accepted at CoRL 2023, we thank our reviewers for the insightful comments and feedback.
SayTap is also featured in our Google AI blog:
See you in Atlanta :)
SayTap: Language to Quadrupedal Locomotion
We use foot contact patterns as interface to bridge instructions in NL and low-level control commands.
New paper w/ Wenhao Yu, Jie Tan,
@heiga_zen
,
@AleksandraFaust
,
@ttyharada
Web
PDF
3
25
92
0
6
30
↓ Work by my amazing colleagues
Introducing a new language-to-reward system for interfacing LLMs with robots using reward functions. Learn how the system’s predictive control tool enables users to teach robots novel actions using natural language inputs →
34
332
1K
0
8
29
Thanks for tweeting about our work ;)
We have more results at
SayTap: Language to Quadrupedal Locomotion
paper page:
Large language models (LLMs) have demonstrated the potential to perform high-level planning. Yet, it remains a challenge for LLMs to comprehend low-level commands, such as joint angle targets or
0
42
161
0
2
26
Me: I need to finish some work these days, the deadlines are tight
Wife: Maybe you don't need to. We'll know that in a couple of weeks
Me: ......
5
0
23
Our robot not only understands direct instructions such as “trot forward fast” but also responds to vague human commands.
Its reaction to “Act as if the ground is very hot” is well aligned with my expectation and is my personal favorite.
Check out our website for more videos!
0
1
23
Join our VLM+🤖 workshop at
#ICRA
in Yokohama, Japan! Share your cool projects, meet awesome people, and explore cutting-edge research. The paper submission deadline is March 11 🚀
Large Language Models (LLMs) and Vision-Language Models (VLMs) are poised to revolutionize robotics.
Join our workshop at
#ICRA2024
on VLMs/LLMs for scene understanding, decision making, control, and more:
Submissions due March 11, 2024!
4
31
144
0
5
16
We contribute a LLM prompt, a reward design and a training pipeline that allows us to train a quadrupedal locomotion controller in ~20 min on a single V100 GPU. And the controller can be transferred to a real robot without any fine-tuning.
Figure↓ gives an overview of our system
1
2
16
What he said ↓
Join us today to chat about our
#NeurIPS2021
paper “The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning” at the spotlight session
info
poster
08:30PST, 11:30EST, 16:30GMT, 01:30JST
0
21
122
0
0
13
🚀 Just released: Our NeuroEvoBench is here to evolve your research. It's like a digital Darwinism dojo! 🥋
🎉 Stoked to share NeuroEvoBench – a JAX-based Evolutionary Optimizer benchmark for Deep Learning 🦎/🧬
🌎 To be presented at
#NeurIPS2023
Datasets & Benchmarks with
@yujin_tang
&
@alanyttian
🌐:
📜:
🧑💻:
5
24
117
0
1
12
While gradient descent has been very successful, let's not forget there are other options that may lead to surprisingly good results. This work by
@RobertTLange
is super interesting!
🦎/🧬Learned Evolutionary Optimization (& Rob 😋) are going on tour! Super excited to be giving talks about our recent work on meta-discovering attention-based ES/GA & JAX during the coming days 🎙️
@AutomlSeminar
: Today 4pm CET
@ml_collective
: Tomorrow 7pm CET
Come & say hi 🤗
2
14
87
0
0
13
I love how Goten and Trunks' fusion in Dragon Ball exemplifies that 1+1 can sometimes be more than 2
1
1
12
Fusing Gundam with Transformer is the "Gotenks" magic in the model space!
By merging a JP LM with specialist models, we achieve top perf on various JP LM benchmarks.
We explored param and data flow space merging, and will further propel the tech across modalities and functions.
Sakana AIの最初の研究成果である、進化的計算による基盤モデル構築に関する論文を公開しました。多様な既存モデルを自動的に融合し優れた基盤モデルを構築するための方法を提案すると共に、それにより試作したモデルを公開しました。
ブログ
論文
15
569
2K
0
1
12
Vizier is an indispensable tool for many ML engineers inside Google, I'm sure external users will benefit from it as well. Also, I'm glad that our EvoJAX () is used by this open source Vizier.
Oh sweet, Google's internal black-box optimization algorithm suite is now open source:
16
280
2K
0
2
11
At each step, an agent can choose to stay still or move to one of the N,S,E,W neighbors. It also observes the entire field. We employ an attention based policy network and PGPE to solve this task. The learnt policy zero-shot generalizes to different number of agents↓
0
0
10
We simplify the new intrinsic reward with BH inequality to make training tractable. We also introduce a discriminative model that learns to tell genuine transitions from fake trajectories for better embeddings.
Paper:
Video:
0
0
8
This summarizes well what I'm telling my friends every time they visit.
2
0
8
Flight to CoRL has been delayed for 7 hours, I'm gonna miss the connecting flight... I hope United can work out a solution for me
3
0
6
X, I agree with them, but you can do better with recommendations!
0
0
7
@dshihlai
@hardmaru
@kazunori_279
@alanyttian
@gcloudpartners
@GoogleAI
In terms of accuracy, we are providing an estimated solution, but scalability is also an advantage in addition to speed.
0
0
6
DEIR scales novelty-based intrinsic reward with a conditional mutual information term that relates actions with distances between past and present observations. Agents are thus able to tell true novelties from those rooted from the stochasticity in the environments' dynamics.
1
0
6
This is so cool!
Introducing 𝗥𝗼𝗯𝗼𝗣𝗶𝗮𝗻𝗶𝘀𝘁 🎹🤖, a new benchmark for high-dimensional robot control! Solving it requires mastering the piano with two anthropomorphic hands.
This has been one year in the making, and I couldn’t be happier to release it today! Some highlights below:
54
235
1K
0
3
6
I'd like to thank
@moverfitted
for taking the time and effort in making the video, it is the highest possible reward for the authors. I especially love the extra comparison results presented near the end of the video, I wish we've done it.
The Sensory Neuron as a Transformer in PyTorch via
@YouTube
Definitely a cool paper and I hope some of you could find my take on the implementation helpful. I only focused on the CartPoleSwingUp task to make things easier.
@yujin_tang
@hardmaru
0
0
7
0
1
5
0
0
4
@noguchis
Thanks for trying our models! For the 10B model, can you try loading it with bfloat16? I believe that's what caused the inference speed difference.
1
0
4
My first thought (before reading the text): an image generating system created this with prompts like "cooking ramen with burning trees". Now I wonder what images will be created with "a tree hit by lightning exposes its vascular system", I guess they won't be close to this
0
0
4
@moverfitted
@hardmaru
Yes :)
I can't give an exact date now, but I promise to release the code
1
0
3
0
0
3
Read this book at the weekend. PFN is a great company, this book tells about its vision, value, management. I like chap 8 the most where it emphasizes personal robots. I'm also enthusiastic about robots, and hope what I'm doing can add to this bright future.
0
0
3
Google cares about mental health, there are sessions that introduce methods to deal with stress. One important thing is to get good sleeps. My personal experience says the most effective way to have sound sleeps is to NEVER check your agent's learning curve before going to bed.
1
0
3
@danbri
@hardmaru
Not quite. We don't train the agent with permuted inputs and hope it memorize various patterns, PI is entirely due to design. On the other hand, in occluded Pong, we did drop some fraction of inputs. But it's more general than dropout, we can accept inputs of arbitrary sizes.
1
0
3
```
import subprocess as sp
p1 = sp.Popen(['echo', s], stdout=sp.PIPE)
p2 = sp.Popen(['xargs'], stdin=p1.stdout)
```
Probably a No Hire, haha
Python interview question:
Remove all leading & trailing spaces WITHOUT using built-in String methods!
How would you implement it?
142
57
501
0
0
2
@truth_tesla
@karpathy
@hardmaru
The agent has two parts: self-att based visual module and a controller. The former processes the entire image (quote: maximize input data), the latter uses only the selected patches' feature (quote: throw away data based on training). Is this different from what you meant?
0
0
2
Nice abstract art via evolution strategy!
Modern Evolution Strategies for Creativity: Fitting Concrete Images and Abstract Concepts
It's a fun work w/
@hardmaru
Pdf
Web
Code
9
88
423
0
0
1
@Troika_London
Thanks to
@zzznah
's pure jax renderer (), I was able to run entire training pipeline on accelerators.
This is only an initial result and there can be a lot of improvements. I'll release the code as an EvoJAX example.
0
1
2
@rishabh16_
@hardmaru
@risi1979
has successfully trained a world model of millions weights using GA. Maybe you can try that:
1
0
2
Meanwhile, JPY is ...
Russia’s ruble hit a seven-year high, cementing its status as the world’s best-performing currency.
It has gained about 35% so far this year, and has more than doubled from a low after the invasion of Ukraine.
897
1K
3K
0
0
2
And it doesn't take millions of rollouts
0
0
2
😂🤣
To be fair, this is hard for human (me) too
0
0
2
“A learning-based locomotion controller enables a quadrupedal ANYmal robot to traverse challenging natural environments.”
0
0
2
He's actually good. Wonder who's going to be the successor.
BREAKING: Shinzo Abe, Japan's longest-serving prime minister, is expected to step down due to health
38
451
650
0
0
2
@hardmaru
Assuming they collect labelled 💩 dataset, they might be able to train a dog health check system too
0
0
2
@RoySnapir
@FlorinGogianu
@hardmaru
@nt_duong
@googlejapan
Hi, you can find the code here:
I’ll also push some updates in the next week or so
1
0
2
😂🤣
Shinjiro Koizumi, son of former PM Junichiro Koizumi, is widely disliked in Japan for dumb statements that have zero substance.
As a result, he's the source of many memes. People often just post a picture of his face with a meaningless or obvious statement underneath. Thread.
106
1K
7K
1
0
2
@FilipoGiovanni
@hardmaru
The controller receives the position of each patch as its input, instead of the content.
0
0
1
Not sure this can be solved in the life time, but challenge accepted.
0
0
1
This would be an interesting yet very challenging task for robots.
0
0
1
😂😂
My worse nightmare would be if Jonny’s mom and mine are friends.
501
28K
161K
0
0
1
Biotech 101 for computer scientists
does a great job breaking down what is in the Pfizer vaccine.
15
248
632
0
0
1
🙇💪
They are surprisingly strong. Fitness exercises for the elderly in parks in Beijing.💪💪💪😅😅😅
150
3K
7K
0
0
0
@tkasasagi
Oasis is great for in the bath/train reading (scribe is a little heavy and not water proof). For papers, you'll need a larger screen. I have one Fujitsu Quaderno, and borrowed a boox from a friend, both are great. I don't take notes, so I have no experience with their pens.
1
0
1