paddle fliud1.3 fluid.ParallelExecutor killed
Created by: shouonga
if args.do_train: print("start exec_strategy!!!") exec_strategy = fluid.ExecutionStrategy() exec_strategy.use_experimental_executor = args.use_fast_executor exec_strategy.num_threads = 4 exec_strategy.num_iteration_per_drop_scope = args.num_iteration_per_drop_scope train_exe = fluid.ParallelExecutor( #运行到此处直接被killed use_cuda=args.use_cuda, loss_name=loss.name, exec_strategy=exec_strategy, main_program=train_program) #train_exe = fluid.Executor(fluid.CPUPlace()) train_pyreader.decorate_tensor_provider(train_data_generator)
只显示killed,没有任何别的错误