I got error on mjc_peg_images

[spk921]

I am using TF v0.12 and got following error when I try to run mjc_peg_images. Also I tried with TFv0.11 same error. Which version of TF works ?

DEBUG:No gps_agent_pkg: gps_agent_pkg ROS path [0]=/opt/ros/kinetic/share/ros ROS path [1]=/opt/ros/kinetic/share I tensorflow/stream_executor/dso_loader.cc:125] successfully opened CUDA library libcublas.so.8.0 locally I tensorflow/stream_executor/dso_loader.cc:125] successfully opened CUDA library libcudnn.so.5 locally I tensorflow/stream_executor/dso_loader.cc:125] successfully opened CUDA library libcufft.so.8.0 locally I tensorflow/stream_executor/dso_loader.cc:125] successfully opened CUDA library libcuda.so.1 locally I tensorflow/stream_executor/dso_loader.cc:125] successfully opened CUDA library libcurand.so.8.0 locally W tensorflow/core/platform/cpu_feature_guard.cc:95] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations. Traceback (most recent call last): File "python/gps/gps_main.py", line 410, in main() File "python/gps/gps_main.py", line 395, in main gps = GPSMain(hyperparams.config, args.quit) File "python/gps/gps_main.py", line 51, in init self.algorithm = config['algorithm']'type' File "python/gps/algorithm/algorithm_badmm.py", line 34, in init self._hyperparams['policy_opt'], self.dO, self.dU File "python/gps/algorithm/policy_opt/policy_opt_tf.py", line 47, in init self.init_network() File "python/gps/algorithm/policy_opt/policy_opt_tf.py", line 71, in init_network network_config=self._hyperparams['network_params']) File "python/gps/algorithm/policy_opt/tf_model_example.py", line 147, in multi_modal_network 'bc2': init_bias([num_filters[1]]), File "python/gps/algorithm/policy_opt/tf_model_example.py", line 13, in init_bias return tf.get_variable(name, initializer=tf.zeros(shape, dtype='float')) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 987, in get_variable custom_getter=custom_getter) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 889, in get_variable custom_getter=custom_getter) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 347, in get_variable validate_shape=validate_shape) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 332, in _true_getter caching_device=caching_device, validate_shape=validate_shape) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.py", line 638, in _get_single_variable name, "".join(traceback.format_list(tb)))) ValueError: Variable None already exists, disallowed. Did you mean to set reuse=True in VarScope? Originally defined at:

File "python/gps/algorithm/policy_opt/tf_model_example.py", line 13, in init_bias return tf.get_variable(name, initializer=tf.zeros(shape, dtype='float')) File "python/gps/algorithm/policy_opt/tf_model_example.py", line 146, in multi_modal_network 'bc1': init_bias([num_filters[0]]), File "python/gps/algorithm/policy_opt/policy_opt_tf.py", line 71, in init_network network_config=self._hyperparams['network_params'])

INFO:signal_shutdown [atexit]

[cbfinn]

It was written and tested for v0.5-0.8.

[foolyc]

add name for the init_bias biases = { 'bc1': init_bias([num_filters[0]], name='bc1'), 'bc2': init_bias([num_filters[1]], name='bc2'), } but you will still faces other bugs when run the mjc_peg_images in tf0.12

[fhoekstra]

I'm getting this same error on Tensorflow 0.8.0 (on Ubuntu)

@foolyc Should adding the name also be necessary for tensorflow 0.8 ? As @cbfinn says it was tested for v0.5-0.8

EDIT: easy try, it fixes the error on Tensorflow 0.8.0 but now I'm wondering if I will find many other bugs due to TF 0.8 not being supported? I thought it was but maybe I understood 0.5-0.8 incorrectly

FYI I am modifying the box2d agent to use images as input. I do not have Mujoco installed, but got the same error when trying to use the multi modal network.