Verified90% confidenceFactTime unknown
Triton is a GPU programming language developed by OpenAI.
10
Sources
90%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Unsloth：本地微调大模型速度提升5倍的开源神器
githubunslothai
Related Entities
Related Claims
UnverifiedTriton是OpenAI开源的GPU编程语言与编译器，其JIT缓存机制将编译产物存储在~/.triton/cache目录，命中缓存则跳过编译84% similarUnverifiedUnsloth's performance improvements come from hand-written GPU kernels using OpenAI's Triton language.76% similarVerifiedTriton Inference Server is an open-source model serving framework from NVIDIA that supports dynamic batching and concurrent model execution, and is compatible with inference backends such as TensorRT, PyTorch, TensorFlow, and ONNX Runtime.69% similarUnverifiedUnsloth采用OpenAI开发的开源GPU编程语言Triton重写了反向传播中的关键算子，包括交叉熵损失计算、RoPE位置编码等68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/33216API
curl https://kongchang.com/api/v1/knowledge/claims/33216MCP
get_claim(id=33216)