Support for zero-dimensional tensors is broken

arogozhnikov commented 4 years ago

Description

Error Message

(Paste the complete error message. Please also include stack trace by setting environment variable DMLC_LOG_STACK_TRACE_DEPTH=10 before running your script.)

terminate called after throwing an instance of 'dmlc::Error'
  what():  [06:19:35] src/imperative/./imperative_utils.h:146: Operator _zeros inferring shapes failed.
input shapes:
output shapes:
None
operator attributes:
dtype : float32
ctx : cpu(0)
shape : []

Stack trace:
  [bt] (0) /usr/local/lib/python3.7/site-packages/mxnet/libmxnet.so(+0x307d3b) [0x7f3ee1e63d3b]
  [bt] (1) /usr/local/lib/python3.7/site-packages/mxnet/libmxnet.so(mxnet::imperative::SetShapeType(mxnet::Context const&, nnvm::NodeAttrs const&, std::vector<mxnet::NDArray*, std::allocator<mxnet::NDArray*> > const&, std::vector<mxnet::NDArray*, std::allocator<mxnet::NDArray*> > const&, mxnet::DispatchMode*)+0x363b) [0x7f3ee4fff3bb]

To Reproduce

It can appear in multiple different situations:

import mxnet
x = mxnet.nd.zeros([]) # failure

or

x = mxnet.nd.from_numpy(numpy.zeros([])) # ok, tensor created
x.attach_grad() # failure

Environment

----------Python Info----------
Version      : 3.7.3
Compiler     : GCC 6.3.0 20170516
Build        : ('default', 'Jun 11 2019 01:05:09')
Arch         : ('64bit', '')
------------Pip Info-----------
Version      : 19.1.1
Directory    : /usr/local/lib/python3.7/site-packages/pip
----------MXNet Info-----------
Version      : 1.6.0
Directory    : /usr/local/lib/python3.7/site-packages/mxnet
Num GPUs     : 0
Commit Hash   : 6eec9da55c5096079355d1f1a5fa58dcf35d6752
----------System Info----------
Platform     : Linux-4.19.76-linuxkit-x86_64-with-debian-9.9
system       : Linux
node         : 0a353fd2b50d
release      : 4.19.76-linuxkit
version      : #1 SMP Thu Oct 17 19:31:58 UTC 2019
----------Hardware Info----------
machine      : x86_64
processor    :
Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                4
On-line CPU(s) list:   0-3
Thread(s) per core:    1
Core(s) per socket:    1
Socket(s):             4
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 158
Model name:            Intel(R) Core(TM) i9-8950HK CPU @ 2.90GHz
Stepping:              10
CPU MHz:               2900.000
BogoMIPS:              5808.00
L1d cache:             32K
L1i cache:             32K
L2 cache:              256K
L3 cache:              12288K
Flags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht pbe syscall nx pdpe1gb lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq dtes64 ds_cpl ssse3 sdbg fma cx16 xtpr pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch pti fsgsbase bmi1 hle avx2 bmi2 erms rtm xsaveopt arat
----------Network Test----------
Setting timeout: 10
Timing for MXNet: https://github.com/apache/incubator-mxnet, DNS: 0.0286 sec, LOAD: 0.7676 sec.
Timing for GluonNLP GitHub: https://github.com/dmlc/gluon-nlp, DNS: 0.0016 sec, LOAD: 0.6134 sec.
Timing for GluonNLP: http://gluon-nlp.mxnet.io, DNS: 0.0497 sec, LOAD: 0.7350 sec.
Timing for D2L: http://d2l.ai, DNS: 0.0560 sec, LOAD: 0.2496 sec.
Timing for D2L (zh-cn): http://zh.d2l.ai, DNS: 0.0302 sec, LOAD: 0.1806 sec.
Timing for FashionMNIST: https://repo.mxnet.io/gluon/dataset/fashion-mnist/train-labels-idx1-ubyte.gz, DNS: 0.1056 sec, LOAD: 0.3243 sec.
Timing for PYPI: https://pypi.python.org/pypi/pip, DNS: 0.0205 sec, LOAD: 0.5132 sec.
Error open Conda: https://repo.continuum.io/pkgs/free/, HTTP Error 403: Forbidden, DNS finished in 0.017886877059936523 sec.

leezu commented 4 years ago

This only works in the https://github.com/apache/incubator-mxnet/issues/14253 interface, which is still opt-in but will be default in 2.0

arogozhnikov commented 4 years ago

@leezu tried with mxnet.numpy.zeros([]) but also got failure.

Both hands for that proposal, BTW.

sxjscience commented 4 years ago

@arogozhnikov You will need to manually call mx.npx.set_np() to trigger the numpy semantic.

sxjscience commented 4 years ago

@arogozhnikov I’ll close it for now because I think adding “mx.npx.set_np()” will solve the issue. Feel free to reopen it.

apache / mxnet