[CPU error] running ppo_snake.py on CPU/CentOS 8 raises error

[anthony0727]

might be :

• change in newer torch version

(py37) [anthony@localhost]~/rl2/rl2/examples% python3 ppo_snake.py
/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/autograd/__init__.py:147: UserWarning: Error detected in AddmmBackward. Traceback of forward call that caused the error:
  File "ppo_snake.py", line 118, in <module>
    log_dir = train(config)
  File "ppo_snake.py", line 64, in train
    worker.run()
  File "/home/anthony/rl2/rl2/workers/base.py", line 146, in run
    done, info, results = self.rollout()
  File "/home/anthony/rl2/rl2/workers/base.py", line 87, in rollout
    info_a = self.agent.step(self.obs, ac, rew, done, obs)
  File "/home/anthony/rl2/rl2/agents/ppo.py", line 217, in step
    info = self.train(advs)
  File "/home/anthony/rl2/rl2/agents/ppo.py", line 242, in train
    loss = self.loss_func(batch_data, self.model, hidden=hidden)
  File "/home/anthony/rl2/rl2/agents/ppo.py", line 23, in loss_func
    ac_dist, val_dist = model(obs, hidden=hidden, mask=dones)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/anthony/rl2/rl2/models/torch/base.py", line 113, in inner
    results = func(self, *args, **kwargs)
  File "/home/anthony/rl2/rl2/agents/ppo.py", line 102, in forward
    action_dist = self.policy(obs, **kwargs)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/anthony/rl2/rl2/models/torch/base.py", line 360, in forward
    output = self.head(ir)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/anthony/rl2/rl2/distributions/torch/distributions.py", line 60, in forward
    x = self.linear(x)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/anthony/rl2/rl2/networks/torch/networks.py", line 19, in forward
    return self.body(x)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/container.py", line 119, in forward
    input = module(input)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/modules/linear.py", line 94, in forward
    return F.linear(input, self.weight, self.bias)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/nn/functional.py", line 1753, in linear
    return torch._C._nn.linear(input, weight, bias)
 (Triggered internally at  /pytorch/torch/csrc/autograd/python_anomaly_mode.cpp:104.)
  allow_unreachable=True, accumulate_grad=True)  # allow_unreachable flag
model saved in PPOAgent_20210607224131/ckpt/8k
[TUTORIAL|ERROR] 2021-06-07 22:41:31,681 > RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [64, 5]], which is output 0 of TBackward, is at version 4; expected version 3 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!
[TUTORIAL|DEBUG] 2021-06-07 22:41:31,681 > Traceback (most recent call last):
  File "ppo_snake.py", line 118, in <module>
    log_dir = train(config)
  File "ppo_snake.py", line 64, in train
    worker.run()
  File "/home/anthony/rl2/rl2/workers/base.py", line 146, in run
    done, info, results = self.rollout()
  File "/home/anthony/rl2/rl2/workers/base.py", line 87, in rollout
    info_a = self.agent.step(self.obs, ac, rew, done, obs)
  File "/home/anthony/rl2/rl2/agents/ppo.py", line 217, in step
    info = self.train(advs)
  File "/home/anthony/rl2/rl2/agents/ppo.py", line 244, in train
    self.model.value.step(loss)
  File "/home/anthony/rl2/rl2/models/torch/base.py", line 381, in step
    loss.backward(retain_graph=retain_graph)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/tensor.py", line 245, in backward
    torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
  File "/home/anthony/.pyenv/versions/3.7.10/envs/py37/lib/python3.7/site-packages/torch/autograd/__init__.py", line 147, in backward
    allow_unreachable=True, accumulate_grad=True)  # allow_unreachable flag
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [64, 5]], which is output 0 of TBackward, is at version 4; expected version 3 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!