I am following ‘Compile Keras Models’ (https://tvm.apache.org/docs/tutorials/frontend/from_keras.html). Instead of gpu, I set up target as ‘llvm’ and ctx as cpu. Then, I executed the following code. It takes about 12,000 millseconds. The warning message is 'Cannot find config for ‘dense_nopack.x86’ and ‘conv2d_NCHWc.x86’. Is there any solution to make better performance? I noticed that similar issue was generated one year ago, but no follow up.
target = ‘llvm’ ctx = tvm.cpu(0) with tvm.transform.PassContext(opt_level=3): executor = relay.build_module.create_executor(‘graph’, mod, ctx, target) dtype = ‘float32’ tvm_out = executor.evaluate()(tvm.nd.array(data.astype(dtype)), **params)