I tried to deploy ssd with gluoncv according the tvm docs. I ran the source code file. The error infos as below.
Traceback (most recent call last):
File "deploy_ssd_gluoncv.py", line 105, in <module>
graph, lib, params = compile(target)
File "deploy_ssd_gluoncv.py", line 80, in compile
net, params = relay.frontend.from_mxnet(block, {"data": dshape})
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 612, in from_mxnet
sym = _from_mxnet_impl(sym, shape, dtype)
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 504, in _from_mxnet_impl
jgraph = json.loads(symbol.tojson())
And here is the related code in mxnet.py:
if isinstance(symbol, mx.sym.Symbol):
params = {}
arg_params = arg_params if arg_params else {}
aux_params = aux_params if aux_params else {}
for k, v in arg_params.items():
params[k] = _nd.array(v.asnumpy())
for k, v in aux_params.items():
params[k] = _nd.array(v.asnumpy())
shape, dtype = _update_shape_dtype(shape, dtype, params)
sym = _from_mxnet_impl(symbol, shape, dtype)
elif isinstance(symbol, mx.gluon.HybridBlock):
if arg_params is not None or aux_params is not None:
raise ValueError("arg_params and aux_params ae not used when importing HybridBlock")
params = {}
for k, v in symbol.collect_params().items():
params[k] = _nd.array(v.data().asnumpy())
data = mx.sym.Variable("data")
sym = symbol(data)
shape, dtype = _update_shape_dtype(shape, dtype, params)
sym = _from_mxnet_impl(sym, shape, dtype)
def _from_mxnet_impl(symbol, shape_dict, dtype_info):
assert symbol is not None
jgraph = json.loads(symbol.tojson())
jnodes = jgraph["nodes"]
node_map = {}
the root of this error is that sym is a tuple instead of symbol. It seems like there is a minor mistake when fetching symbol from hybrid block model. But i don’t know how to fix it.
my envs:
macOS: 10.14.3
mxnet-mkl : 1.3
gluoncv: 0.3
tvm built from source commit id: fd9fa4bf
I modified serveal lines in mxnet.py to fetch symbol from hybrid block according to Quick tip: converting Gluon models to symbolic format.
I replaced
data = mx.sym.Variable("data")
sym = symbol(data)
with
import numpy as np
import mxnet as mx
model_name = "ssd_512_resnet50_v1_voc"
symbol.hybridize()
x = np.zeros([1, 3, 512, 512])
x = mx.nd.array(x)
symbol.forward(x)
symbol.export(model_name)
sym, _, _ = mx.model.load_checkpoint(model_name, 0)
the exception of tojson is fixed, but another error raised:
/Users/cxt123/Applications/anaconda3/envs/conda36_base/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
from ._conv import register_converters as _register_converters
[09:46:18] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 147456 bytes with malloc directly
[09:46:18] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 589824 bytes with malloc directly
[09:46:18] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 2359296 bytes with malloc directly
[09:46:18] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 9437184 bytes with malloc directly
Traceback (most recent call last):
File "deploy_ssd_gluoncv.py", line 94, in <module>
graph, lib, params = compile(target)
File "deploy_ssd_gluoncv.py", line 70, in compile
net, params = relay.frontend.from_mxnet(block, {"data": dshape})
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 625, in from_mxnet
sym = _from_mxnet_impl(sym, shape, dtype)
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 521, in _from_mxnet_impl
res = _convert_map[op_name](children, attrs)
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 199, in _mx_slice_axis
shape = ir_pass.infer_type(inputs[0]).checked_type.shape
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/ir_pass.py", line 45, in infer_type
return _ir_pass.infer_type(expr, mod)
File "/Users/cxt123/Reposities/tvm/python/tvm/_ffi/_ctypes/function.py", line 185, in __call__
ctypes.byref(ret_val), ctypes.byref(ret_tcode)))
File "/Users/cxt123/Reposities/tvm/python/tvm/_ffi/base.py", line 71, in check_call
raise TVMError(py_str(_LIB.TVMGetLastError()))
tvm._ffi.base.TVMError: [09:46:19] /Users/cxt123/Reposities/tvm/src/relay/ir/error.cc:112:
Error(s) have occurred. We have annotated the program with them:
The detail error message as below:
/Users/cxt123/Applications/anaconda3/envs/conda36_base/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
from ._conv import register_converters as _register_converters
[09:54:53] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 147456 bytes with malloc directly
[09:54:53] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 589824 bytes with malloc directly
[09:54:53] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 2359296 bytes with malloc directly
[09:54:53] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 9437184 bytes with malloc directly
Traceback (most recent call last):
File "deploy_ssd_gluoncv.py", line 94, in <module>
graph, lib, params = compile(target)
File "deploy_ssd_gluoncv.py", line 70, in compile
net, params = relay.frontend.from_mxnet(block, {"data": dshape})
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 625, in from_mxnet
sym = _from_mxnet_impl(sym, shape, dtype)
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 521, in _from_mxnet_impl
res = _convert_map[op_name](children, attrs)
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/frontend/mxnet.py", line 199, in _mx_slice_axis
shape = ir_pass.infer_type(inputs[0]).checked_type.shape
File "/Users/cxt123/Reposities/tvm/python/tvm/relay/ir_pass.py", line 45, in infer_type
return _ir_pass.infer_type(expr, mod)
File "/Users/cxt123/Reposities/tvm/python/tvm/_ffi/_ctypes/function.py", line 185, in __call__
ctypes.byref(ret_val), ctypes.byref(ret_tcode)))
File "/Users/cxt123/Reposities/tvm/python/tvm/_ffi/base.py", line 71, in check_call
raise TVMError(py_str(_LIB.TVMGetLastError()))
tvm._ffi.base.TVMError: [09:54:54] /Users/cxt123/Reposities/tvm/src/relay/ir/error.cc:112:
Error(s) have occurred. We have annotated the program with them:
In `main`:
fn () {
free_var %data: Tensor[(1, 3, 512, 512), float32]
free_var %ssd0_resnetv10_conv0_weight: Tensor[(64, 3, 7, 7), float32]
%0 = nn.conv2d(%data, %ssd0_resnetv10_conv0_weight, strides=[2, 2], padding=[3, 3], channels=64, kernel_size=[7, 7]) #
free_var %ssd0_resnetv10_batchnorm0_gamma: Tensor[(64,), float32]
free_var %ssd0_resnetv10_batchnorm0_beta: Tensor[(64,), float32]
free_var %ssd0_resnetv10_batchnorm0_running_mean: Tensor[(64,), float32]
free_var %ssd0_resnetv10_batchnorm0_running_var: Tensor[(64,), float32]
%1 = nn.batch_norm(%0, %ssd0_resnetv10_batchnorm0_gamma, %ssd0_resnetv10_batchnorm0_beta, %ssd0_resnetv10_batchnorm0_running_mean, %ssd0_resnetv10_batchnorm0_running_var) #
%2 = %1.0
%3 = nn.relu(%2) #
%4 = nn.max_pool2d(%3, pool_size=[3, 3], strides=[2, 2], padding=[1, 1]) #
free_var %ssd0_resnetv10_stage1_conv0_weight: Tensor[(64, 64, 1, 1), float32]
...
...
%340 = add(%321, 18f) #
%341 = add(%321, 19f) #
%342 = (%322, %323, %324, %325, %326, %327, %328, %329, %330, %331, %332, %333, %334, %335, %336, %337, %338, %339, %340, %341)
%343 = concatenate(%342, axis=-1) #
%344 = ones_like(%343) #
%345 = multiply(%344, -1f) #
%346 = where(%319, %343, %345) # an internal invariant was violdated whiletypechecking your program[09:54:54] /Users/cxt123/Reposities/tvm/src/relay/op/tensor/transform.cc:965: Check failed: reporter->AssertEQ(cond_shape[i], x_shape[i]) Shape of condition [1, 6132, 21] must be either equal to x or has dimension of 1.
Stack trace returned 10 entries:
[bt] (0) 0 libtvm.dylib 0x000000010cc45140 dmlc::StackTrace(unsigned long) + 464
[bt] (1) 1 libtvm.dylib 0x000000010cc44e24 dmlc::LogMessageFatal::~LogMessageFatal() + 52
[bt] (2) 2 libtvm.dylib 0x000000010d171965 tvm::relay::WhereRel(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&) + 2629
[bt] (3) 3 libtvm.dylib 0x000000010d0800cf void tvm::runtime::detail::unpack_call_dispatcher<bool, 0, 4, bool (*)(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&)>::run<tvm::runtime::TVMArgValue, tvm::runtime::TVMArgValue, tvm::runtime::TVMArgValue, tvm::runtime::TVMArgValue>(bool (* const&)(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&), tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*, tvm::runtime::TVMArgValue&&, tvm::runtime::TVMArgValue&&, tvm::runtime::TVMArgValue&&, tvm::runtime::TVMArgValue&&) + 95
[bt] (4) 4 libtvm.dylib 0x000000010d080029 std::__1::__function::__func<void tvm::runtime::TypedPackedFunc<bool (tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&)>::AssignTypedLambda<bool (*)(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&)>(bool (*)(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&))::'lambda'(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*), std::__1::allocator<void tvm::runtime::TypedPackedFunc<bool (tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&)>::AssignTypedLambda<bool (*)(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&)>(bool (*)(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&))::'lambda'(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*)>, void (tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>::operator()(tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 137
[bt] (5) 5 libtvm.dylib 0x000000010d28bd3d tvm::TypedEnvFunc<bool (tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&)>::operator()(tvm::Array<tvm::relay::Type, void> const&, int, tvm::Attrs const&, tvm::relay::TypeReporter const&) const + 365
[bt] (6) 6 libtvm.dylib 0x000000010d28b665 tvm::relay::TypeSolver::Solve() + 1125
[bt] (7) 7 libtvm.dylib 0x000000010d26e7a4 tvm::relay::TypeInferencer::Infer(tvm::relay::Expr) + 116
[bt] (8) 8 libtvm.dylib 0x000000010d26f948 tvm::relay::InferType(tvm::relay::Function const&, tvm::relay::Module const&, tvm::relay::GlobalVar const&) + 472
[bt] (9) 9 libtvm.dylib 0x000000010d036f08 tvm::relay::ModuleNode::Add(tvm::relay::GlobalVar const&, tvm::relay::Function const&, bool) + 152
;
%346
}
Stack trace returned 10 entries:
[bt] (0) 0 libtvm.dylib 0x000000010cc45140 dmlc::StackTrace(unsigned long) + 464
[bt] (1) 1 libtvm.dylib 0x000000010cc44e24 dmlc::LogMessageFatal::~LogMessageFatal() + 52
[bt] (2) 2 libtvm.dylib 0x000000010cffe32e tvm::relay::ErrorReporter::RenderErrors(tvm::relay::Module const&, bool) + 5406
[bt] (3) 3 libtvm.dylib 0x000000010d26e7c0 tvm::relay::TypeInferencer::Infer(tvm::relay::Expr) + 144
[bt] (4) 4 libtvm.dylib 0x000000010d26f948 tvm::relay::InferType(tvm::relay::Function const&, tvm::relay::Module const&, tvm::relay::GlobalVar const&) + 472
[bt] (5) 5 libtvm.dylib 0x000000010d036f08 tvm::relay::ModuleNode::Add(tvm::relay::GlobalVar const&, tvm::relay::Function const&, bool) + 152
[bt] (6) 6 libtvm.dylib 0x000000010d03870f tvm::relay::ModuleNode::FromExpr(tvm::relay::Expr const&, tvm::Map<tvm::relay::GlobalVar, tvm::relay::Function, void, void> const&) + 943
[bt] (7) 7 libtvm.dylib 0x000000010d26f3c9 tvm::relay::InferType(tvm::relay::Expr const&, tvm::relay::Module const&) + 793
[bt] (8) 8 libtvm.dylib 0x000000010d28954f std::__1::__function::__func<tvm::relay::$_1, std::__1::allocator<tvm::relay::$_1>, void (tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)>::operator()(tvm::runtime::TVMArgs&&, tvm::runtime::TVMRetValue*&&) + 159
[bt] (9) 9 libtvm.dylib 0x000000010d4d5266 TVMFuncCall + 70
