How to print VTA instructions in the integration examples?

I have recently started looking at TVM and VTA and would like to see the instructions that are generated for each convolutional layer.

I started by running test_benchmark_topi_conv2d.py example and later modified the code to print debugging information according to this answer (the rest of the code is not touched):

# I added the following line
with vta.build_config(debug_flag=2):
    if "vta" in target.keys:
        mod = vta.build(s, [data, kernel, bias, res],
                        target=target,
                        target_host=env.target_host,
                        name="conv2d")
    else:
        mod = tvm.build(s, [data, kernel, bias, res],
                        target=target,
                        target_host=env.target_host,
                        name="conv2d")

However, running the new code returns the following error:

module 'tvm.make' has no attribute 'stmt_seq'

I would really appreciate it if you let me know how to print debugging information.

1 Like