Tensor conversion error when trying to use TF-Agents with Anytrading
jimdog99 opened this issue · comments
I have been working on a tf_agent test program using latest anytrading environment.
Once I got past the py_environment compatibility issues and got it to load correctly, I ran into the following problem...
I have a line of code in the replaybuffer that does the following: time_step = environment.current_time_step()
But I get back the following error when this is run against the anytrading environment...
File "/home/michael/test9.py", line 196, in timestamp_data
time_step = environment.current_time_step()
File "/usr/local/lib/python3.10/dist-packages/tf_agents/environments/tf_environment.py", line 196, in current_time_step
return self._current_time_step()
File "/usr/local/lib/python3.10/dist-packages/tf_agents/environments/tf_py_environment.py", line 240, in _current_time_step
outputs = tf.numpy_function(
File "/usr/local/lib/python3.10/dist-packages/tensorflow/python/util/traceback_utils.py", line 153, in error_handler
raise e.with_traceback(filtered_tb) from None
File "/usr/local/lib/python3.10/dist-packages/tensorflow/python/framework/constant_op.py", line 102, in convert_to_eager_tensor
return ops.EagerTensor(value, ctx.device_name, dtype)
ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type Positions).
This seems to be related to the Positions class which holds the current and past Positions. Any ideas?
Using latest tensorflow 2.14 with tf-agents 0.18 on python 3.10 with gymnasium 0.29.1
Could you provide a minimal code to reproduce the problem?
Sorry I resolved the issue.... I had a mistake in my TFPyEnvironment which I was using to wrap your Gym.
Please close this issue.