Unverified50% confidenceFactExact time
TorchInductor是基于OpenAI Triton的代码生成后端，Triton的后端正在扩展到AMD ROCm和Intel GPU
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/7/2026
First Seen
Valid until: 12/6/2026
Sources
Related Entities
Related Claims
UnverifiedInductor 后端针对 CPU 生成 OpenMP 代码，针对 GPU 生成优化的 Triton kernel73% similarUnverifiedNVIDIA Triton Inference Server和TorchServe均内置了成熟的动态批处理调度器71% similarUnverifiedtorch.compile的inductor后端会针对每种独特输入张量形状生成专用的融合Triton kernel67% similarUnverifiedAMD于2016年推出ROCm开源GPU计算平台，定位为英伟达CUDA的开源替代方案67% similarUnverifiedAMD通过ROCm构建开源替代方案，并向PyTorch贡献代码使MI300X等GPU获得一流支持65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/870408API
curl https://kongchang.com/api/v1/knowledge/claims/870408MCP
get_claim(id=870408)