⚡ Open Model Pretraining Masterclass — Elie Bakouch, HuggingFace SmolLM 3, FineWeb, FinePDF
Today *Elie Bakouch,* who leads pre-training efforts at Hugging Face and is a key architect behind SmolLM, walks us through his five pillars of model training: data quality optimization, model architecture design, information extraction efficiency, gradient quality maximization, and training stability at scale. We also talked about their open science data work like FineWeb-Edu2 and FinePDF dataset, new Adam optimizer alternatives like Muon and Shampoo, and the evolution of Mixture of Experts (MoE) architectures. Elie broke down recent innovations from DeepSeek's granular routing mechanisms to …
Watch on YouTube ↗
(saves to browser)
Chapters (10)
Introduction
1:10
Hugging Face Research Team Overview
4:20
The Unified View of Model Training
10:17
Optimizer Innovation: Beyond Adam
21:15
MoE Architecture Deep Dive
29:26
Expert Specialization and Routing
33:09
Sparsity Trends and Production Models
40:20
Data Quality and Rephrasing Revolution
43:20
Small LM Training Insights
53:56
Open Source Tools and Future Directions
Playlist
Uploads from Latent Space · Latent Space · 0 of 60
← Previous
Next →
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
![[LLM Paper Club] Llama 3.1 Paper: The Llama Family of Models](https://i.ytimg.com/vi/TgLSYIBoX5U/mqdefault.jpg)
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
![[Paper Club] 🍓 On Reasoning: Q-STaR and Friends!](https://i.ytimg.com/vi/Y5-FeaFOEFM/mqdefault.jpg)
54
![[Paper Club] Writing in the Margins: Chunked Prefill KV Caching for Long Context Retrieval](https://i.ytimg.com/vi/VHwrhL_MSV4/mqdefault.jpg)
55
56
57
58
59
60
Ep 18: Petaflops to the People — with George Hotz of tinycorp
Latent Space
FlashAttention-2: Making Transformers 800% faster AND exact
Latent Space
RWKV: Reinventing RNNs for the Transformer Era
Latent Space
Generating your AI Media Empire - with Youssef Rizk of Wondercraft.ai
Latent Space
RAG is a hack - with Jerry Liu of LlamaIndex
Latent Space
The End of Finetuning — with Jeremy Howard of Fast.ai
Latent Space
Why AI Agents Don't Work (yet) - with Kanjun Qiu of Imbue
Latent Space
Powering your Copilot for Data - with Artem Keydunov from Cube.dev
Latent Space
Beating GPT-4 with Open Source Models - with Michael Royzen of Phind
Latent Space
The State of Silicon and the GPU Poors - with Dylan Patel of SemiAnalysis
Latent Space
The "Normsky" architecture for AI coding agents — with Beyang Liu + Steve Yegge of SourceGraph
Latent Space
The AI-First Graphics Editor - with Suhail Doshi of Playground AI
Latent Space
The Accidental AI Canvas - with Steve Ruiz of tldraw
Latent Space
The Origin and Future of RLHF: the secret ingredient for ChatGPT - with Nathan Lambert
Latent Space
The Four Wars of the AI Stack - Dec 2023 Recap
Latent Space
The State of AI in production — with David Hsu of Retool
Latent Space
Building an open AI company - with Ce and Vipul of Together AI
Latent Space
Truly Serverless Infra for AI Engineers - with Erik Bernhardsson of Modal
Latent Space
A Brief History of the Open Source AI Hacker - with Ben Firshman of Replicate
Latent Space
Open Source AI is AI we can Trust — with Soumith Chintala of Meta AI
Latent Space
Making Transformers Sing - with Mikey Shulman of Suno
Latent Space
A Comprehensive Overview of Large Language Models - Latent Space Paper Club
Latent Space
Why Google failed to make GPT-3 -- with David Luan of Adept
Latent Space
Personal AI Meetup - Bee, BasedHardware, LangChain LangFriend, Deepgram EmilyAI
Latent Space
Supervise the Process of AI Research — with Jungwon Byun and Andreas Stuhlmüller of Elicit
Latent Space
Breaking down the OG GPT Paper by Alec Radford
Latent Space
High Agency Pydantic over VC Backed Frameworks — with Jason Liu of Instructor
Latent Space
This World Does Not Exist — Joscha Bach, Karan Malhotra, Rob Haisfield (WorldSim, WebSim, Liquid AI)
Latent Space
LLM Asia Paper Club Survey Round
Latent Space
How to train a Million Context LLM — with Mark Huang of Gradient.ai
Latent Space
How AI is Eating Finance - with Mike Conover of Brightwave
Latent Space
How To Hire AI Engineers (ft. James Brady and Adam Wiggins of Elicit)
Latent Space
State of the Art: Training 70B LLMs on 10,000 H100 clusters
Latent Space
The 10,000x Yolo Researcher Metagame — with Yi Tay of Reka
Latent Space
Training Llama 2, 3 & 4: The Path to Open Source AGI — with Thomas Scialom of Meta AI
Latent Space
[LLM Paper Club] Llama 3.1 Paper: The Llama Family of Models
Latent Space
Synthetic data + tool use for LLM improvements 🦙
Latent Space
RLHF vs SFT to break out of local maxima 📈
Latent Space
The Winds of AI Winter (Q2 Four Wars of the AI Stack Recap)
Latent Space
Segment Anything 2: Memory + Vision = Object Permanence — with Nikhila Ravi and Joseph Nelson
Latent Space
Answer.ai & AI Magic with Jeremy Howard
Latent Space
Is finetuning GPT4o worth it?
Latent Space
Personal benchmarks vs HumanEval - with Nicholas Carlini of DeepMind
Latent Space
Building AGI with OpenAI's Structured Outputs API
Latent Space
Q* for model distillation 🍓
Latent Space
Finetuning LoRAs on BILLIONS of tokens 🤖
Latent Space
Cursor UX team is CRACKED 💻
Latent Space
Choosing the BEST OpenAI model 🏆
Latent Space
How will OpenAI voice mode change API design?
Latent Space
Ephemeral software FOR YOU 💻
Latent Space
STEALING OpenAI models data 🥷
Latent Space
Models can do MORE than what you ask for 🤖
Latent Space
[Paper Club] 🍓 On Reasoning: Q-STaR and Friends!
Latent Space
[Paper Club] Writing in the Margins: Chunked Prefill KV Caching for Long Context Retrieval
Latent Space
The Ultimate Guide to Prompting - with Sander Schulhoff from LearnPrompting.org
Latent Space
llm.c's Origin and the Future of LLM Compilers - Andrej Karpathy at CUDA MODE
Latent Space
Prompt Engineer is NOT a job 📝
Latent Space
Prompt Mining LLMs for better prompts ⛏️
Latent Space
The six pillars of few-shot prompting 🔧
Latent Space
Language Agents: From Reasoning to Acting — with Shunyu Yao of OpenAI, Harrison Chase of LangGraph
Latent Space
DeepCamp AI