mirror of
https://github.com/wassname/ray.git
synced 2026-07-10 02:57:21 +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>
101 lines
3.7 KiB
Python
101 lines
3.7 KiB
Python
import logging
|
|
import numpy as np
|
|
|
|
from ray.rllib.models.torch.torch_modelv2 import TorchModelV2
|
|
from ray.rllib.models.torch.misc import SlimFC, normc_initializer
|
|
from ray.rllib.utils.annotations import override
|
|
from ray.rllib.utils.framework import get_activation_fn
|
|
from ray.rllib.utils import try_import_torch
|
|
|
|
_, nn = try_import_torch()
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
class FullyConnectedNetwork(TorchModelV2, nn.Module):
|
|
"""Generic fully connected network."""
|
|
|
|
def __init__(self, obs_space, action_space, num_outputs, model_config,
|
|
name):
|
|
TorchModelV2.__init__(self, obs_space, action_space, num_outputs,
|
|
model_config, name)
|
|
nn.Module.__init__(self)
|
|
|
|
activation = get_activation_fn(
|
|
model_config.get("fcnet_activation"), framework="torch")
|
|
hiddens = model_config.get("fcnet_hiddens")
|
|
no_final_linear = model_config.get("no_final_linear")
|
|
|
|
# TODO(sven): implement case: vf_shared_layers = False.
|
|
# vf_share_layers = model_config.get("vf_share_layers")
|
|
|
|
logger.debug("Constructing fcnet {} {}".format(hiddens, activation))
|
|
layers = []
|
|
prev_layer_size = np.product(obs_space.shape)
|
|
self._logits = None
|
|
|
|
# Create layers 0 to second-last.
|
|
for size in hiddens[:-1]:
|
|
layers.append(
|
|
SlimFC(
|
|
in_size=prev_layer_size,
|
|
out_size=size,
|
|
initializer=normc_initializer(1.0),
|
|
activation_fn=activation))
|
|
prev_layer_size = size
|
|
|
|
# The last layer is adjusted to be of size num_outputs, but it's a
|
|
# layer with activation.
|
|
if no_final_linear and self.num_outputs:
|
|
layers.append(
|
|
SlimFC(
|
|
in_size=prev_layer_size,
|
|
out_size=self.num_outputs,
|
|
initializer=normc_initializer(1.0),
|
|
activation_fn=activation))
|
|
prev_layer_size = self.num_outputs
|
|
# Finish the layers with the provided sizes (`hiddens`), plus -
|
|
# iff num_outputs > 0 - a last linear layer of size num_outputs.
|
|
else:
|
|
if len(hiddens) > 0:
|
|
layers.append(
|
|
SlimFC(
|
|
in_size=prev_layer_size,
|
|
out_size=hiddens[-1],
|
|
initializer=normc_initializer(1.0),
|
|
activation_fn=activation))
|
|
prev_layer_size = hiddens[-1]
|
|
if self.num_outputs:
|
|
self._logits = SlimFC(
|
|
in_size=hiddens[-1],
|
|
out_size=self.num_outputs,
|
|
initializer=normc_initializer(0.01),
|
|
activation_fn=None)
|
|
else:
|
|
self.num_outputs = (
|
|
[np.product(obs_space.shape)] + hiddens[-1:-1])[-1]
|
|
|
|
self._hidden_layers = nn.Sequential(*layers)
|
|
|
|
# TODO(sven): Implement non-shared value branch.
|
|
self._value_branch = SlimFC(
|
|
in_size=prev_layer_size,
|
|
out_size=1,
|
|
initializer=normc_initializer(1.0),
|
|
activation_fn=None)
|
|
# Holds the current value output.
|
|
self._cur_value = None
|
|
|
|
@override(TorchModelV2)
|
|
def forward(self, input_dict, state, seq_lens):
|
|
obs = input_dict["obs_flat"]
|
|
features = self._hidden_layers(obs.reshape(obs.shape[0], -1))
|
|
logits = self._logits(features) if self._logits else features
|
|
self._cur_value = self._value_branch(features).squeeze(1)
|
|
return logits, state
|
|
|
|
@override(TorchModelV2)
|
|
def value_function(self):
|
|
assert self._cur_value is not None, "must call forward() first"
|
|
return self._cur_value
|