training issue about dsrnn

[hoonm1n]

Hi @Shuijing725 !.
I have a issue about your code in training with srnn.
I changed robot.policy in config.py from "selfAttn_merge_srnn" to "srnn".
Then when i run the train.py, the error occurred like under line.

`<Monitor<CrowdSimPredRealGST>>
No ghost version.
new gst
new st model
LOADED MODEL
device: cuda:0

Traceback (most recent call last):
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/train.py", line 247, in
main()
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/train.py", line 90, in main
actor_critic = Policy(
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/rl/networks/model.py", line 28, in init
self.base = base(obs_shape, base_kwargs)
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/rl/networks/srnn_model.py", line 378, in init
robot_size = 7 if args.env_type == 'crowd_sim' else 2
AttributeError: 'Namespace' object has no attribute 'env_type'. Did you mean: 'env_name'`?

Could you solve this issue? Thank you for reading!

[hoonm1n]

When i modified that line in srnn_model.py, another error occurred like under line.
Traceback (most recent call last):
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/train.py", line 247, in
main()
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/train.py", line 162, in main
value, action, action_log_prob, recurrent_hidden_states = actor_critic.act(
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/rl/networks/model.py", line 60, in act
value, actor_features, rnn_hxs = self.base(inputs, rnn_hxs, masks, infer=True)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/root/dsrnn_ws/CrowdNav_Prediction_AttnGraph/rl/networks/srnn_model.py", line 441, in forward
nodes_current_selected = self.robot_linear(robot_node)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/linear.py", line 114, in forward
return F.linear(input, self.weight, self.bias)
RuntimeError: mat1 and mat2 shapes cannot be multiplied (16x7 and 2x3)

[Shuijing725]

Do you want to add trajectory prediction? If not, change env_name in arguments.py to CrowdSimVarNum-v0.
Otherwise, S-RNN is not compatible with prediction right now, so I'll have to fix it.

[hoonm1n]

Does that mean that the old network's weights can't be trained in the current repository(CrowdNav_Prediction_AttnGraph) yet?

[Shuijing725]

This repo has several different gym environments (see readme -> training -> ii). S-RNN can be trained with CrowdSimVarNum-v0 environment, which does not have human trajectory prediction. But it cannot be trained with other environments with prediction.