serving icon indicating copy to clipboard operation
serving copied to clipboard
verified publisher icon tensorflow

TensorFlow serving on RTX 5090D

Open reindeergao opened this issue 8 months ago • 0 comments

I use latest docker mirror tensorflow/serving:2.19.0-gpu on the Nvidia RTX 5090D The following error was reported:

W external/org_tensorflow/tensorflow/core/framework/op_kernel.cc:1829] INTERNAL: 'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY, blockZ, 0, reinterpret_cast<CUstream>(stream), params, nullptr)' failed with 'CUDA_ERROR_INVALID_HANDLE' W external/org_tensorflow/tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleLoadData(&module, data)' failed with 'CUDA_ERROR_INVALID_PTX' W external/org_tensorflow/tensorflow/compiler/mlir/tools/kernel_gen/tf_gpu_runtime_wrappers.cc:40] 'cuModuleGetFunction(&function, module, kernel_name)' failed with 'CUDA_ERROR_INVALID_HANDLE'

Could a tensorflow_model_server or docker image that supports Nvidia RTX 5090D be provided?

Thanks!!!

reindeergao avatar Jun 19 '25 08:06 reindeergao