[Bug] [Relay] squeeze shape function does not transform axis

[Lucien0]

I have this Relay Program:

import numpy as np
import tvm
from tvm import relay, transform, IRModule

x_shape = (1, relay.Any(), 100, 17, 1)

x = relay.var('x', shape=x_shape)
y = relay.squeeze(x, axis=[-1])
y = relay.sqrt(y)
mod = IRModule.from_expr(y)
with transform.PassContext(opt_level=1):
    vm_exec = relay.vm.compile(mod, target='llvm')
vm = tvm.runtime.vm.VirtualMachine(vm_exec, tvm.cpu())

x_input = np.zeros(shape=(1, 100, 100, 17, 1), dtype='float32')
data = vm.run(x=x_input)

Expected behavior

run correctly.

Actual behavior

raise error:
Check failed: ret == 0 (-1 vs. 0) : Assert fail: 4 == T.tvm_struct_get(arg_T_sqrt, 0, 4, "int32"), arg.T_sqrt.ndim is expected to equal 4

Environment

None

Steps to reproduce

run this case

I find in squeeze shape function:

tvm/python/tvm/relay/op/_transform.py

Lines 914 to 927 in f21a17b

	@_reg.register_shape_func("squeeze", False)
	def squeeze_shape_func(attrs, inputs, _):
	"""
	Shape function for squeeze op.
	"""
	axis = attrs.axis if attrs.axis is None else get_const_tuple(attrs.axis)
	keep_axes = []
	remove_axes = []
	if axis is not None:
	for i in range(inputs[0].shape[0].value):
	if i not in axis:
	keep_axes.append(i)
	else:
	remove_axes.append(i)

Here if the axis is -1, then will get the wrong result.

cc @shingjan