chintantrivedi / rl-bot-football Goto Github PK
View Code? Open in Web Editor NEWAn RL agent for the Google Football environment
An RL agent for the Google Football environment
Hello
Thanks for the detailed implementation.. but I get 0 rewards. And consequently, no model is saved.
When calling the fit function on the actor model, Im getting the following error when executing your complete code (with image_based = True):
ValueError: Data cardinality is ambiguous:
x sizes: 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 72, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 128, 128, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
y sizes: 128
Please provide data which shares the same first dimension.
Hi, @ChintanTrivedi I am using the modified version of your code to train the environment created using the Unity engine.
[I have modified the code to handle this].
Action space = Continuous
Observation space = Vectorized
TensorFlow version = 2.3.0
Python version = 3.8.3
I think the error is introduced due to the custom loss function call. [Tensor is passed to the loss function so eager execution is giving error]
So while I am trying to fit the model to the collected data, I am getting the following errors.
So, please can you tell me what am I doing wrong?
Epoch 1/8
Traceback (most recent call last):
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\execute.py", line 59, in quick_execute
tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
TypeError: An op outside of the function building code is being passed
a "Graph" tensor. It is possible to have Graph tensors
leak out of the function building context by including a
tf.init_scope in your function building code.
For example, the following function will fail:
@tf.function
def has_init_scope():
my_constant = tf.constant(1.)
with tf.init_scope():
added = my_constant * 2
The graph tensor has name: old_prediction:0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "temp.py", line 259, in <module>
agent.train()
File "temp.py", line 228, in train
actor_loss = self.actor.fit(
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\keras\engine\training.py", line 108, in _method_wrapper
return method(self, *args, **kwargs)
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1098, in fit
tmp_logs = train_function(iterator)
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\def_function.py", line 780, in __call__
result = self._call(*args, **kwds)
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\def_function.py", line 840, in _call
return self._stateless_fn(*args, **kwds)
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\function.py", line 2829, in __call__
return graph_function._filtered_call(args, kwargs) # pylint: disable=protected-access
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\function.py", line 1843, in _filtered_call
return self._call_flat(
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\function.py", line 1923, in _call_flat
return self._build_call_outputs(self._inference_function.call(
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\function.py", line 545, in call
outputs = execute.execute(
File "C:\Users\dhyey\Desktop\Train-ml-agents\python-envs\offline_training\lib\site-packages\tensorflow\python\eager\execute.py", line 72, in quick_execute
raise core._SymbolicException(
tensorflow.python.eager.core._SymbolicException: Inputs to eager execution function cannot be Keras symbolic tensors, but found [<tf.Tensor 'old_prediction:0' shape=(None, 2) dtype=float32>, <tf.Tensor 'advantage:0' shape=(None, 1) dtype=float32>, <tf.Tensor 'reward:0' shape=(None, 1) dtype=float32>, <tf.Tensor 'value:0' shape=(None, 1) dtype=float32>]
Here is the code which is calling the .fit() method to the model:
while not target_reached and iters < max_iters:
states = []
actions = []
old_predictions = []
rewards = []
values = []
masks = []
state = None
self.env.reset()
step_result = self.env.get_steps(self.behavior_name)
state = step_result[0].obs[0] # observation
for itr in range(ppo_steps):
#generate the actions using predict method on actor model.
action, action_matrix, predicted_action = self.get_action(state, True)
q_value = self.critic.predict(state, steps=1)
next_state, reward, done = self.step(action) # apply the actions to the env and returns the env response.
print('itr: ' + str(itr) + ', reward=' + str(reward) + ', q_val=' + str(q_value))
mask = not done
states.append(state)
actions.append(action_matrix)
old_predictions.append(predicted_action)
rewards.append(reward)
values.append(q_value)
masks.append(mask)
state = next_state
if done:
self.env.reset()
q_value = self.critic.predict(state, steps=1)
values.append(q_value)
returns, advantages = self.get_advantages(values, masks, rewards)
# reshaping
states = np.reshape(states, (len(states), self.state_dims))
actions = np.reshape(actions, (len(actions), self.n_actions))
old_predictions = np.reshape(old_predictions, (len(old_predictions), self.n_actions))
rewards = np.reshape(rewards, (-1, 1))
values = np.reshape(values, (len(values), 1))
advantages = np.reshape(advantages, (len(advantages), 1))
returns = np.reshape(returns, (len(returns), 1))
actor_loss = self.actor.fit(
[states, old_predictions, advantages, rewards, values[:-1]],
[actions],
verbose=True, shuffle=True, epochs=8,
callbacks=[self.tensor_board]
)
critic_loss = self.critic.fit(
[states], [returns],
shuffle=True, epochs=8, verbose=True,
callbacks=[self.tensor_board]
)
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.