Daniel Han
where else to find them
what they talk about
where they keep going back
- PyTorch Conference2 editions2024–2025
- AI Engineer World's Fair2 editions2024–2025
on a stage with
never shared a talk
who talks about the same things
- Chenyang Zhao3 talks
- Legare Kerrison3 talks
- Gabriel Preda2 talks
- Alessandro Cappelli2 talks
- John Davies6 talks
- Lutske de Leeuw6 talks
| title | conference | date |
|---|---|---|
| Maximizing Luck in Reinforcement Learning | PyTorch Conference 2025 | 2025-10-22 |
Unsloth claims RL training speeds and memory drop by specific tweaks: float8 quantization and keeping the LM head in float32 for accuracy, plus algorithm swaps… | ||
| Can Reinforcement Learning Lead to AGI? | AI Infra Summit at PyTorch Conference 2025 | 2025-10-21 |
Claims that RL can scale to AGI rest on concrete pivots: scaling inference (currently ~90% of compute), automating reward functions, and handling long context… | ||
| [Full Workshop] Reinforcement Learning, Kernels, Reasoning, Quantization & Agents | AI Engineer World's Fair 2025 | 2025-06-03 |
Claims that RL can replace stages of SFT, plus a concrete algorithmic tweak: GRPO removes the value model and uses rollout statistics as the baseline, letting… | ||
| Hacks to Make LLM Training Faster | PyTorch Conference 2024 | 2024-09-18 |
Six practical wins: switching from float32 to float16/BF16 yields roughly 5x speedups on common GPUs; quantizing to int8 keeps accuracy while int4 halves model… | ||
| Low Level Technicals of LLMs | AI Engineer World's Fair 2024 | 2024-06-25 |
Hands-on workshop shows exactly what was fixed and how: debugged Gemma and Phi‑3 implementations (including a rope float32 fix that causes the largest… | ||
| nothing matches | ||