SGLang-native diffusion transformer overrides converted with NVIDIA ModelOpt.
AI & ML interests
LLM, distributed systems
Recent Activity
A collection of production-grade draft models for speculative decoding
-
lmsys/SGLang-EAGLE3-Llama-3.3-70B-Instruct-SpecForge
1B • Updated • 370 -
lmsys/SGLang-EAGLE3-Llama-3.1-8B-Instruct-SpecForge
0.4B • Updated • 700 -
lmsys/SGLang-EAGLE3-Qwen3-30B-A3B-Instruct-2507-SpecForge-Nex
0.2B • Updated • 1.4k • 4 -
lmsys/SGLang-EAGLE3-Llama-4-Scout-17B-16E-Instruct-SpecForge
0.8B • Updated • 37
Train Eagle 3 for SGLang with SpecForge
SGLang-native diffusion transformer overrides converted with NVIDIA ModelOpt.
A collection of production-grade draft models for speculative decoding
-
lmsys/SGLang-EAGLE3-Llama-3.3-70B-Instruct-SpecForge
1B • Updated • 370 -
lmsys/SGLang-EAGLE3-Llama-3.1-8B-Instruct-SpecForge
0.4B • Updated • 700 -
lmsys/SGLang-EAGLE3-Qwen3-30B-A3B-Instruct-2507-SpecForge-Nex
0.2B • Updated • 1.4k • 4 -
lmsys/SGLang-EAGLE3-Llama-4-Scout-17B-16E-Instruct-SpecForge
0.8B • Updated • 37
Train Eagle 3 for SGLang with SpecForge