Kathryn (Jinqi) Chen

ML-systems | music | philosophy | @SCSatCMU

[Tianqi Chen @tqchenml]

📢Excited to introduce Apache TVM FFI, an open ABI and FFI for ML systems, enabling compilers, libraries, DSLs, and frameworks to naturally interop with each other. Ship one library across pytorch, jax, cupy etc and runnable across python, c++, rust https://t.co/m2gHJRreol

[sehoonkim @sehoonkim418]

How can we make LLM agents work together efficiently on complex tasks at a large scale?

🚨Introducing LLMCompiler🦙🛠️, a tool that compiles an effective plan for executing multiple tasks in parallel.

It helps create scalable LLM applications, identifies tasks for parallel execution, and manages dependencies. LLMCompiler is compatible with both open-source and OpenAI models, marking a stride towards more efficient and intelligent software systems. 🧵1/n

📌Link to Paper: https://t.co/q2X1qztJxi
📌Link to Code: https://t.co/zUTxlAW5tT

Joint work with: @snrpsnr @ryan_tabrizi @nicholaszlee Michael Mahoney @KurtKeutzer @amir__gholami

[Werner Vogels @Werner]

Internships at #AWS can lead to quite some impact. This excellent paper by Zhuang Wang and team, on in-memory checkpointing of large ML model training was presented at SOSP this week. (paper pdf: https://t.co/iPMB1DcDFP)
/at @amazonscience: https://t.co/Ypmi8fpg7p

[Stas Bekman @StasBekman]

Boom! @MSFTDeepSpeed implemented Sequence Parallelism (SP)

https://t.co/53vmTj4fqw

For very long sequences the paper shows that DeepSpeed-Ulysses trains 2.5x faster with 4x longer sequence length than the existing method SOTA baseline.

Even if it's not faster, it's super-useful to have this feature available along with ZeRO++!

So let's see which frameworks support SP:

- DeepSpeed - check
- Megatron-LM - check
- Colossal-AI - check
- PyTorch - no yet - cc: @wanchao_

[Peyman Milanfar @docmilanfar]

After I read this brilliant essay about free will, I had no choice by to post it.

[Omar Khattab @lateinteraction]

🚨Announcing 𝗗𝗦𝗣𝘆, the framework for solving advanced tasks w/ LMs.

Express *any* pipeline as clean, Pythonic control flow.

Just ask DSPy to 𝗰𝗼𝗺𝗽𝗶𝗹𝗲 your modular code into auto-tuned chains of prompts or finetunes for GPT, Llama, and/or T5.🧵

https://t.co/QColxp8atV

[Alessio Fanelli @FanaHOVA]

H100s? 🤨 I'll just run LLaMA2 70B in the browser, thanks!

@tqchenml came on @latentspacepod to talk about the work MLC is doing to enable everyone to run models natively on any hardware / software stack, including Chrome and iPhones (and AMD cards!)

🎙️ https://t.co/lPuLBpr1Rq

[Kathryn (Jinqi) Chen @meow_cat_7]

will stay tuned to what consensus AI community can reach on the risks and ethics of AI. also, i believe it's quite necessary for AI technical people to have more conversations with law makers, governors, social workers, linguists, philosophers etc. for more solid understanding.

[Andrew Ng @AndrewYNg]

Had an insightful conversation with @geoffreyhinton about AI and catastrophic risks. Two thoughts we want to share:
(i) It's important that AI scientists reach consensus on risks-similar to climate scientists, who have rough consensus on climate change-to shape good policy.
(ii) Do AI models understand the world? We think they do. If we list out and develop a shared view on key technical questions like this, it will help move us toward consensus on risks.

I learned a lot speaking with Geoff. Let’s all of us in AI keep having conversations to learn from each other!

[Tianqi Chen @tqchenml]

MLC Chat app is now on @AppStore Checkout it out!
Chat with open language models running on your iPad and iPhone, offline, locally, no data collected
https://t.co/CVUuj7OfZi

[Junru Shao @junrushao]

LLMs like RedPajama are embracing open permissive license. What does it mean for us? Could they become our personal friends, and will they shape a unique market? Now RedPajama-3B can run locally on phones, browsers and laptops with hardware acceleration! https://t.co/pJqgEqi6Nd

[Graham Neubig @gneubig]

There are so many chatbots nowadays, it’s hard to keep up!

To help out, we made an open source tool for automatic comparison of chatbots, and created a report on LLaMa, Alpaca, Vicuna, ChatGPT, Cohere, etc.!

Report: https://t.co/n3JfKYcIXB
Browser: https://t.co/kOK4YsNHBq

🧵⬇️

[Chris Olah @ch402]

I was writing an expanded related work section on the relationship between superposition and distributed representations.

... It kind of outgrew being a related work section!

[Bohan Hou @bohanhou1998]

Can LLMs run natively on your iPhone📱? Our answer is yes, and we can do more!

We are introducing MLC-LLM, an open framework that brings language models (LLMs) directly into a broad class of platforms (CUDA, Vulkan, Metal) with GPU acceleration!

Demo: https://t.co/sSmUeGJu6T

[Shashank Prasanna @shshnkp]

New blogpost! a visual primer on how @PyTorch 2.0 compiler technologies for graph capture, IRs, operator fusions and automatic C++ and @NVIDIAAIDev GPU code generation. This is your one stop shop to grok PyTorch 2.0's torch.compile() API Summary 🧵👇

https://t.co/4bpkx3Kuc7

[Hongyi Jin @HongyiJin258]

Introducing WebLLM, an open-source chatbot that brings language models (LLMs) directly onto web browsers. We can now run instruction fine-tuned LLaMA (Vicuna) models natively on your browser tab via @WebGPU with no server support. Checkout our demo at https://t.co/dXII0MzYg1 .

[Kathryn (Jinqi) Chen @meow_cat_7]

@your_fault__ あなたの音��は私の心に響くよ

[Arena.ai @arena]

Introducing Vicuna, an open-source chatbot impressing GPT-4!

🚀 Vicuna reaches 90%* quality of ChatGPT/Bard while significantly outperforming other baselines, according to GPT-4's assessment.

Blog: https://t.co/IfRPy28lDN
Demo: https://t.co/Vzs3OFe5O1

[Zihao Ye @ye_combinator]

Check our #ASPLOS23 presentation tomorrow at 11 in session 4C at Grand D! We’ll be discussing our paper “SparseTIR: Composable Abstractions for Sparse Compilation in Deep Learning” - a compiler for sparse DL workloads built on @ApacheTVM (1/2)

[Yaoyao Ding @YaoyaoDing1]

Excited to share that our paper on "Hidet: Task-Mapping Programming Paradigm for Deep Learning Tensor Programs" has been published in #asplos23. Check out the link for the full paper: https://t.co/Esco7bxFrM. Also, don't miss our talk tomorrow in session 4C at #asplos23!