7 10 29

Tyler Hilbert

TylerHilbert

Tyler-Hilbert

AI & ML interests

Performance Testing

Recent Activity

updated a dataset about 14 hours ago

TylerHilbert/PyTorchConference2025_GithubRepos

replied to pcuenq's post 2 days ago

👉 What happened in AI in 2025? 👈 We prepared the 2025 version of the HF AI Timeline Grid, highlighting open vs API-based model releases, and allowing you to browse and filter by access, modality, and release type! Play with it here: https://huggingface.co/spaces/2025-ai-timeline/2025-ai-timeline Here's my personal quarterly TL;DR: 1️⃣ Q1 — Learning to Reason Deepseek not only releases a top-notch reasoning model, but shows how to train them and compete with closed frontier models. OpenAI debuts Deep Research. Significant milestones: DeepSeek R1 & R1-Zero, Qwen 2.5 VL, OpenAI Deep Research, Gemini 2.5 Pro (experimental) 2️⃣ Q2 — Multimodality and Coding More LLMs embrace multimodality by default, and there's a surge in coding agents. Strong vision, audio, and generative models emerge. Significant milestones: Llama 4, Qwen 3, Imagen 4, OpenAI Codex, Google Jules, Claude 4 3️⃣ Q3 — "Gold" rush, OpenAI opens up, the community goes bananas Flagship models get gold in Math olympiads and hard benchmarks. OpenAI releases strong open source models and Google releases the much anticipated nano-banana for image generation and editing. Agentic workflows become commonplace. Significant milestones: Gemini and OpenAI IMO Gold, gpt-oss, Gemini 2.5 Flash Image, Grok 4, Claude Sonnet 4.5 4️⃣ Q4 — Mistral returns, leaderboard hill-climbing Mistral is back with updated model families. All labs release impressive models to wrap up the year! Significant milestones: Claude Opus 4.5, DeepSeek Math V2, FLUX 2, GPT 5.1, Kimi K2 Thinking, Nano Banana Pro, GLM 4.7, Gemini 3, Mistral 3, MiniMax M2.1 🤯 Credits 🙏 NHLOCAL for the source data https://github.com/NHLOCAL/AiTimeline 🫡 @reach-vb for the original idea, design and recipe 🙌 @ariG23498 and yours truly for compiling and verifying the 2025 edition 🥳 Here's to 2026, wishing it becomes the best year ever for open releases and on-device-first use-cases! 🥂

commented on an article 2 days ago

Fine-Tuning FunctionGemma on TPU to Create a Virtual Fitness Coach in 10 Minutes, $0.50

View all activity

Organizations

None yet

updated a dataset about 14 hours ago

TylerHilbert/PyTorchConference2025_GithubRepos

Viewer • Updated about 14 hours ago • 119 • 106 • 1

replied to pcuenq's post 2 days ago

What AI advancements do you anticipate for 2026?

commented on Fine-Tuning FunctionGemma on TPU to Create a Virtual Fitness Coach in 10 Minutes, $0.50 2 days ago

This is a great tutorial, I particularly like your 3 TPU optimizations.
What other TPU optimization techniques are possible, beyond SPMD Initialization, FSDP and padding to prevent graph recompilation, if you want to further squeeze better performance out of a TPU?

New activity in perplexity-ai/draco 2 days ago

Weights selection

#2 opened 2 days ago by

TylerHilbert

liked a dataset 2 days ago

perplexity-ai/draco

Viewer • Updated about 17 hours ago • 100 • 90 • 26

upvoted a collection 3 days ago

Qwen3-Coder-Next

Collection

4 items • Updated 4 days ago • 65

New activity in allenai/CoSyn-point 3 days ago

NVIDIA Visual Profiler

#1 opened 3 days ago by

TylerHilbert

New activity in siro1/kernelbook-kimi_k2_thinking-synthetic-tasks 5 days ago

Looking for more information about dataset use.

#2 opened 5 days ago by

TylerHilbert

New activity in allenai/SERA-32B 5 days ago

KernelBench

#3 opened 5 days ago by

TylerHilbert

New activity in Qwen/Qwen3-235B-A22B-Instruct-2507 6 days ago

Qwen 3 for performance testing?

#39 opened 6 days ago by

TylerHilbert

liked a dataset 6 days ago

siro1/kernelbook-kimi_k2_thinking-synthetic-tasks

Viewer • Updated 7 days ago • 18.2k • 31 • 2

replied to Javedalam's post 7 days ago

This comment has been hidden

commented on Preference Tuning LLMs with Direct Preference Optimization Methods 7 days ago

I’m curious about your take on applying these to coding-specific models. Since we can automatically verify code, KTO seems like it could bypass the bottleneck of creating preference pairs. Do you think the 'unpaired' nature of KTO makes it the superior choice for technical domains where 'bad' data is easy to generate but 'good' data is expensive?