IBM brings its Spyre accelerator into PyTorch workflows
IBM says its Spyre accelerator now has a PyTorch backend, with a vLLM plugin that connects it to distributed inference through llm-d.
Product
3 stories · Updated
IBM says its Spyre accelerator now has a PyTorch backend, with a vLLM plugin that connects it to distributed inference through llm-d.
Google has released TorchTPU on GitHub, allowing developers to train and serve PyTorch models on its TPUs with minimal code changes, the company says.
Amazon says TorchNeuron lets developers run existing PyTorch code on Trainium and will be open-sourced on October 20.