[BugFix][Runtime] Fix Incorrect node information

[zhaojinxi]

node["attrs"] and node["shape"] may read incorrect node information due to incorrect indexing, especially nodes with multiple outputs in the graph.

[tvm-bot]

Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.

• cc @areusch _{See #10317 for details}

_{Generated by tvm-bot}

[masahi]

Can you come up a test case for this?

[zhaojinxi]

@masahi Try the code below, with the old wrong code, it will print "p0 has shape: [1, 3, 4, 16]", actually it should print "p0 has shape: [3, 3, 1, 1]"

import tvm
import numpy
from tvm import relay
from tvm.contrib.debugger.debug_executor import GraphModuleDebug

x = relay.var("x", shape=(1, 3, 48, 16), dtype="float32")
t = relay.split(x, [12, 16, 32], 2).astuple()
x0 = relay.TupleGetItem(t, 0)
x1 = relay.TupleGetItem(t, 1)
x2 = relay.TupleGetItem(t, 2)
x3 = relay.TupleGetItem(t, 3)
weight = relay.const(numpy.random.uniform(-1, 1, (3, 3, 1, 1)).astype("float32"))
y = relay.nn.conv2d(x2, weight, kernel_size=(1, 1), kernel_layout="OIHW", out_dtype="float32") + x3

func = relay.Function([x], relay.Tuple([x0, x1, y]))
mod = tvm.IRModule.from_expr(func)
mod = relay.transform.InferType()(mod)
target = tvm.target.Target("llvm")
device = tvm.cpu()
lib = relay.build(mod, target=target)
m = GraphModuleDebug(lib["debug_create"]("default", device), [device], lib.get_graph_json(), ".")
nodes = m.debug_datum.get_graph_nodes()
print(nodes[2]['name'],' has shape: ',nodes[2]['shape'])

[masahi]

I mean, can you add that as a test, probably in unittest/test_runtime_graph_debug.py.

[zhaojinxi]

I mean, can you add that as a test, probably in unittest/test_runtime_graph_debug.py.

done