torchdiffeq odeint.py is no longer compatible with pytorch==1.3.1

[VaiSuliafu]

``Hi,

After setting up my environment as indicated on the README, I discovered torchdiffeq._impl.odeint.py contains the following import:

from torch.autograd.functional import vjp

However, torch.autograd.functional doesn't exist in pytorch 1.3.1.

Then, updating torch and torchvision to their latest versions results in the following error:

Traceback (most recent call last): File "run.py", line 137, in <module> oderl.train_env_model(num_iters=args.iters) File "/root/repos/mbrl-smdp-ode/mbrl.py", line 161, in train_env_model train=True) File "/root/repos/mbrl-smdp-ode/model.py", line 581, in compute_loss self.predict_next_states(traj_cur_states, actions, time_steps, lengths, train=train) File "/root/repos/mbrl-smdp-ode/model.py", line 563, in predict_next_states train=train) # [N, T+1, D_latent] File "/root/repos/mbrl-smdp-ode/model.py", line 508, in encode_latent_traj means_z0, logvars_z0 = self.encoder_z0(torch.cat((states, actions), dim=-1), time_steps, lengths) File "/root/miniconda3/envs/mbrl_env/lib/python3.6/site-packages/torch/nn/modules/module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "/root/repos/mbrl-smdp-ode/model.py", line 78, in forward data_packed = pack_padded_sequence(data, lengths, batch_first=True, enforce_sorted=False) File "/root/miniconda3/envs/mbrl_env/lib/python3.6/site-packages/torch/nn/utils/rnn.py", line 245, in pack_padded_sequence _VF._pack_padded_sequence(input, lengths, batch_first) RuntimeError: 'lengths' argument should be a 1D CPU int64 tensor, but got 1D cuda:0 Long tensor (mbrl_env) root@insideDocker:~/repos/mbrl-smdp-ode#

Can anyone help resolve this? Trying to reproduce the results of this paper for baseline measurements of some new work :)

[mqjinwon]

I got the same problem, and I changed the PyTorch version to 1.7.0. It works to me.

here's my conda environment file

name: mbrl
channels:

• defaults
  dependencies:
• _libgcc_mutex=0.1=main
• _openmp_mutex=4.5=1_gnu
• ca-certificates=2021.7.5=h06a4308_1
• certifi=2021.5.30=py36h06a4308_0
• libedit=3.1.20210714=h7f8727e_0
• libffi=3.2.1=hf484d3e_1007
• libgcc-ng=9.3.0=h5101ec6_17
• libgomp=9.3.0=h5101ec6_17
• libstdcxx-ng=9.3.0=hd4cf53a_17
• ncurses=6.2=he6710b0_1
• openssl=1.1.1l=h7f8727e_0
• pip=21.2.2=py36h06a4308_0
• python=3.6.7=h0371630_0
• readline=7.0=h7b6447c_5
• setuptools=58.0.4=py36h06a4308_0
• sqlite=3.33.0=h62c20be_0
• tk=8.6.11=h1ccaba5_0
• wheel=0.37.0=pyhd3eb1b0_1
• xz=5.2.5=h7b6447c_0
• zlib=1.2.11=h7b6447c_3
• pip:
  • cffi==1.14.6
  • cloudpickle==2.0.0
  • cycler==0.10.0
  • cython==0.29.24
  • dataclasses==0.8
  • fasteners==0.16.3
  • future==0.18.2
  • glfw==2.3.0
  • gym==0.20.0
  • imageio==2.9.0
  • kiwisolver==1.3.1
  • matplotlib==3.3.4
  • mujoco-py==2.0.2.13
  • numpy==1.19.5
  • pillow==8.3.2
  • pycparser==2.20
  • pyparsing==2.4.7
  • python-dateutil==2.8.2
  • scipy==1.4.0
  • six==1.16.0
  • torch==1.7.0
  • torchvision==0.8.0
  • typing-extensions==3.10.0.2
  • torchdiffeq==0.2.2
    prefix: "your anaconda environment path"