Tongfei Chen

159 posts

Tongfei Chen

Tongfei Chen

@ctongfei

Researcher in #NLProc; Functional programmer @scala_lang; {Natural | Programming} language enthusiast; NLP/ML/PL. Tweets are my own

Katılım Ağustos 2014
709 Takip Edilen355 Takipçiler
Tongfei Chen retweetledi
Yunmo Chen
Yunmo Chen@YunmoChen·
📢 Exciting News in NLP Metrics! 📊 🔍 There are a lot of different metrics for structured prediction tasks for NLP. Or are there? Our EMNLP 2023 paper suggests they might not be so different at all! Plus, we've got a great new Python library to help you make your own metrics! 🚀
GIF
English
1
10
37
8.7K
Tongfei Chen
Tongfei Chen@ctongfei·
@xtimv @thomasahle Unpopular opinion: All math notations should be 0-indexed. \sum_0^n should be applied to {0, 1, ..., n-1}.
English
0
0
1
44
Tim Vieira
Tim Vieira@xtimv·
@thomasahle I really like [a:b) = [a, a+1,.., b-1] (a:b) = [a+1, .., b-1] (a:b] = [a+1, .., b-1, b] [a:b] = [a, a+1, .., b-1, b] It's the continuous brackets, but the colon makes it discrete. You can also drop either end point to have it be unbounded.
English
3
0
5
290
Thomas Ahle
Thomas Ahle@thomasahle·
Some useful notation
English
2
0
0
1.2K
Marc Marone
Marc Marone@ruyimarone·
@SashaMTL My understanding is that it's not, due to things like sum reductions operating in non deterministic order determined by the hardware at runtime + fp addition not being associative
English
2
0
4
894
Sasha Luccioni, PhD 🦋🌎✨🤗
Sasha Luccioni, PhD 🦋🌎✨🤗@SashaMTL·
Is it possible for the training of an LLM to be fully reproducible? Like down to the resulting weights? If so, what would you need to report? If not, why not? (Real question, real debate)
English
34
13
85
50.7K
小島みなこ 🏳️‍⚧️
小島みなこ 🏳️‍⚧️@MinakoOikawa·
我有点好奇。。。paper 之间会不会有引用关系。。是不是 dag。。。会不会有互相引用的现象。。
中文
8
0
10
87.1K
Tim Vieira
Tim Vieira@xtimv·
@mrdrozdov I think scala would allow a min= b, but I'm not sure about precedence with say a min= b * c. (@ctongfei ?)
English
1
0
0
0
Tim Vieira
Tim Vieira@xtimv·
In Python, we can write 𝚡 += 𝚢, 𝚡 *= 𝚢, 𝚡 **= 𝚢, 𝚡 @= 𝚢, ... for all infix binary operators. When will we be able to write 𝚡 𝚖𝚒𝚗= 𝚢 for any binary function? The status quo 𝚡 = 𝚖𝚒𝚗(𝚡, 𝚢) is so tedious!
English
10
0
33
0
Tongfei Chen
Tongfei Chen@ctongfei·
@xleaps @mrseaoxbleem These optimization techniques (quasi-Newton, trust-region methods, etc) usually require some estimation of the Hessian matrices, which are computationally infeasible for overparameterized neural nets.
English
0
0
0
0
Eric Xu (e/Mettā)
Eric Xu (e/Mettā)@xleaps·
@mrseaoxbleem 我觉得惋惜的是当年有许多的优化技巧和技术，现在全都被 SGD 这种不上台面的算法加上超快的 GPU 给暴力遮盖了。
中文
1
0
5
0
Tongfei Chen
Tongfei Chen@ctongfei·
Wordle 212 5/6 🟩⬛⬛⬛🟩 🟩⬛⬛⬛🟩 🟩⬛🟩⬛🟩 🟩⬛🟩⬛🟩 🟩🟩🟩🟩🟩
English
0
0
0
0
Michael Lin, MD PhD 🧬
Michael Lin, MD PhD 🧬@michaelzlin·
Can someone help me identify this fruit? I've been eating them but it dawned on me that there may not be a randomized control trial demonstrating their long-term safety.
Michael Lin, MD PhD 🧬 tweet media
English
15
2
33
0
Sabrina J. Mielke
Sabrina J. Mielke@sjmielke·
i learned to program with this way way way back when and i miss it so much sometimes
English
2
0
6
0
Tongfei Chen
Tongfei Chen@ctongfei·
Log is a linear operator 😛
English
0
0
3
0
Tongfei Chen
Tongfei Chen@ctongfei·
@keiskS I’m still in Baltimore, but will move to Seattle soon!
English
1
0
0
0
keisks
keisks@keiskS·
@ctongfei Yeah, it’s crazy… stay cool!!
English
1
0
0
0

Keşfet

@martinmbauer @xtimv @thomasahle @ruyimarone @SashaMTL @lidangzzz @MinakoOikawa @rendu_a