| Mobile Web

Search results for PyTorch

AI & Enterprise

AMD, Intel integrate x86 AI software to counter Arm push

AMD and Intel will work together to make AI development on x86 CPUs easier, HotHardware reported. The two companies plan to integrate their PyTorch optimisation technologies into the open-source oneDNN library. The move is expected to let developers use the same software on processors from both companies and reduce the need to optimise AI workloads separately for each CPU. The cooperation is seen as a response to Arm-based competitors.

AI & Enterprise

Google open-sources TorchTPU to help move PyTorch apps to TPUs

Google has released an open-source project called TorchTPU designed to help developers move PyTorch-based AI applications to its TPU chips and Google Cloud. The tool aims to run existing PyTorch apps on TPUs without rewriting code, requiring only simple configuration changes. TechCrunch also noted Amazon has launched TorchNeuron for moving PyTorch apps to its Trainium chips and AWS, while submitting code to decouple features from Nvidia chips.

AI & Enterprise

Amazon open-sources TorchNeuron to run PyTorch code on Trainium without changes

Amazon said it will release TorchNeuron, a tool that lets developers run PyTorch code on its Trainium AI chips without modifying the code. It said developers can switch chips by changing a single setting and continue using existing testing, debugging and scaling tools. Amazon plans to open-source TorchNeuron at the PyTorch Conference North America on Oct. 20. It also said developer time, not chips, is a major cost for companies.