mirror of
https://github.com/wassname/ray.git
synced 2026-06-29 05:34:49 +08:00
Sven Mika
428516056a
* Policy-classes cleanup and torch/tf unification. - Make Policy abstract. - Add `action_dist` to call to `extra_action_out_fn` (necessary for PPO torch). - Move some methods and vars to base Policy (from TFPolicy): num_state_tensors, ACTION_PROB, ACTION_LOGP and some more. * Fix `clip_action` import from Policy (should probably be moved into utils altogether). * - Move `is_recurrent()` and `num_state_tensors()` into TFPolicy (from DynamicTFPolicy). - Add config to all Policy c'tor calls (as 3rd arg after obs and action spaces). * Add `config` to c'tor call to TFPolicy. * Add missing `config` to c'tor call to TFPolicy in marvil_policy.py. * Fix test_rollout_worker.py::MockPolicy and BadPolicy classes (Policy base class is now abstract). * Fix LINT errors in Policy classes. * Implement StatefulPolicy abstract methods in test cases: test_multi_agent_env.py. * policy.py LINT errors. * Create a simple TestPolicy to sub-class from when testing Policies (reduces code in some test cases). * policy.py - Remove abstractmethod from `apply_gradients` and `compute_gradients` (these are not required iff `learn_on_batch` implemented). - Fix docstring of `num_state_tensors`. * Make QMIX torch Policy a child of TorchPolicy (instead of Policy). * QMixPolicy add empty implementations of abstract Policy methods. * Store Policy's config in self.config in base Policy c'tor. * - Make only compute_actions in base Policy's an abstractmethod and provide pass implementation to all other methods if not defined. - Fix state_batches=None (most Policies don't have internal states). * Cartpole tf learning. * Cartpole tf AND torch learning (in ~ same ts). * Cartpole tf AND torch learning (in ~ same ts). 2 * Cartpole tf (torch syntax-broken) learning (in ~ same ts). 3 * Cartpole tf AND torch learning (in ~ same ts). 4 * Cartpole tf AND torch learning (in ~ same ts). 5 * Cartpole tf AND torch learning (in ~ same ts). 6 * Cartpole tf AND torch learning (in ~ same ts). Pendulum tf learning. * WIP. * WIP. * SAC torch learning Pendulum. * WIP. * SAC torch and tf learning Pendulum and Cartpole after cleanup. * WIP. * LINT. * LINT. * SAC: Move policy.target_model to policy.device as well. * Fixes and cleanup. * Fix data-format of tf keras Conv2d layers (broken for some tf-versions which have data_format="channels_first" as default). * Fixes and LINT. * Fixes and LINT. * Fix and LINT. * WIP. * Test fixes and LINT. * Fixes and LINT. Co-authored-by: Sven Mika <sven@Svens-MacBook-Pro.local>
102 lines
4.0 KiB
Python
102 lines
4.0 KiB
Python
from gym.spaces import Discrete, MultiDiscrete, Tuple
|
|
import numpy as np
|
|
from typing import Union
|
|
|
|
from ray.rllib.models.action_dist import ActionDistribution
|
|
from ray.rllib.utils.annotations import override
|
|
from ray.rllib.utils.exploration.exploration import Exploration
|
|
from ray.rllib.utils.framework import try_import_tf, try_import_torch, \
|
|
TensorType
|
|
from ray.rllib.utils.tuple_actions import TupleActions
|
|
from ray.rllib.utils import force_tuple
|
|
|
|
tf = try_import_tf()
|
|
torch, _ = try_import_torch()
|
|
|
|
|
|
class Random(Exploration):
|
|
"""A random action selector (deterministic/greedy for explore=False).
|
|
|
|
If explore=True, returns actions randomly from `self.action_space` (via
|
|
Space.sample()).
|
|
If explore=False, returns the greedy/max-likelihood action.
|
|
"""
|
|
|
|
def __init__(self, action_space, *, model, framework, **kwargs):
|
|
"""Initialize a Random Exploration object.
|
|
|
|
Args:
|
|
action_space (Space): The gym action space used by the environment.
|
|
framework (Optional[str]): One of None, "tf", "torch".
|
|
"""
|
|
super().__init__(
|
|
action_space=action_space,
|
|
model=model,
|
|
framework=framework,
|
|
**kwargs)
|
|
|
|
# Determine py_func types, depending on our action-space.
|
|
if isinstance(self.action_space, (Discrete, MultiDiscrete)) or \
|
|
(isinstance(self.action_space, Tuple) and
|
|
isinstance(self.action_space[0], (Discrete, MultiDiscrete))):
|
|
self.dtype_sample, self.dtype = (tf.int64, tf.int32)
|
|
else:
|
|
self.dtype_sample, self.dtype = (tf.float64, tf.float32)
|
|
|
|
@override(Exploration)
|
|
def get_exploration_action(self,
|
|
*,
|
|
action_distribution: ActionDistribution,
|
|
timestep: Union[int, TensorType],
|
|
explore: bool = True):
|
|
# Instantiate the distribution object.
|
|
if self.framework == "tf":
|
|
return self.get_tf_exploration_action_op(action_distribution,
|
|
explore)
|
|
else:
|
|
return self.get_torch_exploration_action(action_distribution,
|
|
explore)
|
|
|
|
def get_tf_exploration_action_op(self, action_dist, explore):
|
|
def true_fn():
|
|
action = tf.py_function(self.action_space.sample, [],
|
|
self.dtype_sample)
|
|
# Will be unnecessary, once we support batch/time-aware Spaces.
|
|
return tf.expand_dims(tf.cast(action, dtype=self.dtype), 0)
|
|
|
|
def false_fn():
|
|
return tf.cast(
|
|
action_dist.deterministic_sample(), dtype=self.dtype)
|
|
|
|
action = tf.cond(
|
|
pred=tf.constant(explore, dtype=tf.bool)
|
|
if isinstance(explore, bool) else explore,
|
|
true_fn=true_fn,
|
|
false_fn=false_fn)
|
|
|
|
# TODO(sven): Move into (deterministic_)sample(logp=True|False)
|
|
if isinstance(action, TupleActions):
|
|
batch_size = tf.shape(action[0][0])[0]
|
|
else:
|
|
batch_size = tf.shape(action)[0]
|
|
logp = tf.zeros(shape=(batch_size, ), dtype=tf.float32)
|
|
return action, logp
|
|
|
|
def get_torch_exploration_action(self, action_dist, explore):
|
|
if explore:
|
|
# Unsqueeze will be unnecessary, once we support batch/time-aware
|
|
# Spaces.
|
|
a = self.action_space.sample()
|
|
req = force_tuple(
|
|
action_dist.required_model_output_shape(
|
|
self.action_space, self.model.model_config))
|
|
# Add a batch dimension.
|
|
if len(action_dist.inputs.shape) == len(req) + 1:
|
|
a = np.expand_dims(a, 0)
|
|
action = torch.from_numpy(a).to(self.device)
|
|
else:
|
|
action = action_dist.deterministic_sample()
|
|
logp = torch.zeros(
|
|
(action.size()[0], ), dtype=torch.float32, device=self.device)
|
|
return action, logp
|