🌄 I like training interesting transformers.
🌠 The maximum model size I have trained from scratch is 3B, using 128 A100 GPUs. Looking forward to the opportunity to use more GPUs to train larger models in the future!
📫 How to reach me: leo1037987031@gmail.com

Token Usage

Tool	All-time	30d	7d	Top models	Last seen
Codex	`12.64B`	`152.5M`	`19.7M`	`gpt-5.5` `gpt-5.4` `gpt-5.3-codex`	`2026-08-14`
Claude Code	`404.1M`	`0`	`0`	`claude-opus-4-5` `claude-opus-4-6` `gemini-3-pro-high`	`2026-04-23`

Pinned Loading

[ICML 2024 Spotlight] FiT: Flexible Vision Transformer for Diffusion Model

Python 435 14
Forked from xxyQwQ/ComfyBench

[CVPR 2025] ComfyBench: Benchmarking LLM-based Agents in ComfyUI for Autonomously Designing Collaborative AI Systems

Python
[NeurIPS 2023] Sentry-Image: Detect Any AI-generated Images

139 3
Forked from TencentARC/pi-Tuning

[ICML 2023] pi-Tuning: Transferring Multimodal Foundation Models with Optimal Multi-task Interpolation

Python 1
Forked from OpenEarthLab/PredBench

[ECCV 2024] PredBench: Benchmarking Spatio-Temporal Prediction across Diverse Disciplines

Python