I run the example of nnvm coreml in rk3399.When I use the target = “llvm”,the result is right.When I use target = “llvm”, it can’t work. When I use target = “tvm.target.mali()”, it can run, but the result is wrong. Is it that I cannot use the coreml in RK3399’s GPU?
What do you mean by result is wrong ?
Did you follow this tutorial http://nnvm.tvmlang.org/tutorials/deploy_model_on_mali_gpu.html#sphx-glr-tutorials-deploy-model-on-mali-gpu-py ?
I do not follow this tutorial: http://nnvm.tvmlang.org/tutorials/deploy_model_on_mali_gpu.html#sphx-glr-tutorials-deploy-model-on-mali-gpu-py. Because my modek is coreml model so I follow this tutorial: http://nnvm.tvmlang.org/tutorials/from_coreml.html#sphx-glr-tutorials-from-coreml-py.
And I modify “graph, lib, params = nnvm.compiler.build(sym, target, shape_dict, params=params)”,target =tvm.target.mali(). I let ctx = tvm.cl(0).
Then I run, the result is (‘Top-1 id’, 0, ‘class name’, ‘tench, Tinca tinca’).
When I execute set_input(**params), It have these warning.
m.set_input(**params) [09:58:02] src/runtime/graph/graph_runtime.cc:91: Warning: cannot find “batch_norm21_gamma_mul_div_expand” among input [09:58:02] src/runtime/graph/graph_runtime.cc:91: Warning: cannot find “batch_norm10_gamma_mul_div_expand” among input [09:58:02] src/runtime/graph/graph_runtime.cc:91: Warning: cannot find “batch_norm18_add_beta_expand” among input [09:58:02] src/runtime/graph/graph_runtime.cc:91: Warning: cannot find “batch_norm26_add_beta_expand” among input [09:58:02] src/runtime/graph/graph_runtime.cc:91: Warning: cannot find “batch_norm8_gamma_mul_div_expand” among input [09:58:02] src/runtime/graph/graph_runtime.cc:91: Warning: cannot find “batch_norm14_add_beta_expand” among input