Ml Agents release 19 training is freezing. Unity out of memory.

Question

Hi, I'm having trouble training my 5dof robotic arm. I upgraded to release 19 branch, now I get this error and I still run out of memory:

 \ml-agents\mlagents\trainers\torch\networks.py:91: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at  C:\Users\builder\tkoch\workspace\pytorch\pytorch_1647970138273\work\torch\csrc\utils\tensor_new.cpp:201.)
   enc.update_normalization(torch.as_tensor(vec_input))

I don't know why this error comes out now but it may be related to the laggy (10 robot arm training at 19fps in the editor)

here is the config for the trainer:

 [INFO] Hyperparameters for behavior name arm:
         trainer_type:   ppo
         hyperparameters:
           batch_size:   1024
           buffer_size:  10240
           learning_rate:        0.0003
           beta: 0.005
           epsilon:      0.2
           lambd:        0.95
           num_epoch:    7
           learning_rate_schedule:       linear
           beta_schedule:        linear
           epsilon_schedule:     linear
         network_settings:
           normalize:    True
           hidden_units: 32
           num_layers:   2
           vis_encode_type:      simple
           memory:       None
           goal_conditioning_type:       hyper
           deterministic:        False
         reward_signals:
           extrinsic:
             gamma:      0.995
             strength:   1.0
             network_settings:
               normalize:        False
               hidden_units:     128
               num_layers:       2
               vis_encode_type:  simple
               memory:   None
               goal_conditioning_type:   hyper
               deterministic:    False
         init_path:      results\scenario1\arm\checkpoint.pt
         keep_checkpoints:       5
         checkpoint_interval:    500000
         max_steps:      30000000
         time_horizon:   250
         summary_freq:   30000
         threaded:       False
         self_play:      None
         behavioral_cloning:     None

running python 3.7.13.

Anybody have a clue ?

Answer 1

Answer by aymenou190 · Jun 04 at 06:19 PM

Hey, I have sorted out the problem: I had to reinstall PyTorch because I was using the latest version, however mlAgents only supports 1.7.1. To install it, execute the following command in command prompt: torch=1.7.1 pip3 install -f https://download.pytorch.org/whl/torch stable.html

During the installation process, I encountered another problem with Windows not accepting path characters over 260, so I simply followed this guide: https://www.howtogeek.com/266621/how-to-make-windows-10-accept-file-paths-over-260-characters/

Add comment · Share

Ml Agents release 19 training is freezing. Unity out of memory.

1 Reply

Your answer

Follow this Question

Related Questions