def f(preds, values, returns, actions, mask):
advantages = jnp.squeeze(returns - stop_gradient(values), axis=-1)
logps = self._policy_dist.log_prob(preds, actions)
awr_loss = actor_critic.AWRLoss(beta=self._beta, w_max=self._w_max)(
(logps, advantages, jnp.zeros_like(logps), mask))
l2_value_loss = jnp.mean((returns - values)**2) * self._value_loss_coeff
return awr_loss + l2_value_loss
def _do_custom_gradients(self, x, weights, state, rng):
"""Calls this layer for a forward pass, but with custom gradients."""
def _do_forward(y, weights):
old_weights, old_state, old_rng = self._weights, self._state, self._rng
self._weights = weights
res = self.forward(y)
s = self._state
self._weights, self._state, self._rng = old_weights, old_state, old_rng
return res, s
def do_forward_vjp(y, weights):
"""Custom gradient (vjp) function."""
old_weights, old_state, old_rng = self._weights, self._state, self._rng
self._weights = weights
output = self.forward(y)
new_state = self._state
self._weights, self._state, self._rng = old_weights, old_state, old_rng
def vjpfun(grad):
grad = grad[0] # Ignore dummy gradient wrt state.
res = self.backward(y, output, grad, weights, state, new_state,
rng)
return res
return (output, new_state), vjpfun
do_forward = math.custom_grad(do_forward_vjp, _do_forward)
output, state = do_forward(x, weights)
# TODO(lukaszkaiser): Investigate why we need this stop_gradient
state = math.stop_gradient(state)
return output, state
def f(log_probs, advantages, old_log_probs, mask):
if reweight: # Use new policy weights for sampled actions instead.
mask *= jnp.exp(math.stop_gradient(log_probs) - old_log_probs)
if sampled_all_discrete: # Actions were sampled uniformly; weight them.
mask *= jnp.exp(old_log_probs)
weights = jnp.minimum(awr_weights(advantages, beta), w_max)
return -jnp.sum(log_probs * weights * mask) / jnp.sum(mask)
def f(dist_inputs, values, returns, actions, old_log_probs, mask):
"""Definition of the Proximal Policy Optimization loss."""
del mask # TODO(lukaszkaiser): make PPO work with Transformer
ppo_objective = rl_layers.PPOObjective(
dist_inputs,
stop_gradient(values),
returns,
actions,
old_log_probs,
log_prob_fun=self._policy_dist.log_prob,
epsilon=self._epsilon,
normalize_advantages=self._normalize_advantages)
entropy_loss = rl_layers.EntropyLoss(
dist_inputs,
actions,
log_prob_fun=self._policy_dist.log_prob,
entropy_coeff=self._entropy_coeff,
entropy_fun=self._policy_dist.entropy)
l2_value_loss = rl_layers.ValueLoss(
values, returns, value_loss_coeff=self._value_loss_coeff)
return -ppo_objective.mean() + l2_value_loss - entropy_loss
def f(dist_inputs, values, returns, dones, rewards, actions,
old_log_probs, mask):
"""Definition of the A2C loss."""
del dones, rewards, old_log_probs
# Typically we have dist_inputs of the shape float32[128,1,18]
assert len(dist_inputs.shape) == 3, (
f'dist_inputs.shape was {dist_inputs.shape} '
f'but expected length of the tensor shape is 3')
# values of the shape float32[128,1,1]
# returns of the shape float32[128,1,1]
assert values.shape == returns.shape, (
f'values.shape was {values.shape}'
f'returns.shape was (returns.shape)')
# actions of the shape int32[128,1] in the case of discrete actions
# and float32[128,1,6] in the case of of half-cheetah
# actions agree with returns/values on the first two coordinates
assert actions.shape[0:2] == returns.shape[0:2], (
f'actions.shape was {actions.shape}'
f'returns.shape was (returns.shape)')
# and mask of the shape float32[128,1]
assert len(mask.shape) == 2, f'mask.shape was {mask.shape}'
# which agrees with returns/values/actions on the first two coordinates
assert mask.shape[0:2] == returns.shape[0:2], (
f'mask.shape was {mask.shape}'
f'returns.shape was (returns.shape)')
a2c_objective = rl_layers.A2CObjective(
dist_inputs,
stop_gradient(values),
returns,
actions,
mask,
log_prob_fun=self._policy_dist.log_prob,
normalize_advantages=self._normalize_advantages)
# we insist that a2c_objective is a scalar
assert jnp.ndim(
a2c_objective) == 0, f'a2c_objective was {a2c_objective}'
entropy_loss = rl_layers.EntropyLoss(
dist_inputs,
actions,
log_prob_fun=self._policy_dist.log_prob,
entropy_coeff=self._entropy_coeff,
entropy_fun=self._policy_dist.entropy)
assert jnp.ndim(
entropy_loss) == 0, f'entropy_loss was {entropy_loss}'
l2_value_loss = rl_layers.ValueLoss(
values, returns, value_loss_coeff=self._value_loss_coeff)
assert jnp.ndim(
l2_value_loss) == 0, f'l2_value_loss was {l2_value_loss}'
combined_loss = a2c_objective + l2_value_loss - entropy_loss
return combined_loss
def f(dist_inputs, values, returns, actions, old_log_probs, mask):
"""Definition of the Proximal Policy Optimization loss."""
del mask # TODO(lukaszkaiser): make PPO work with Transformer
# We have dist_inputs of the shape float32[128,1,18]
assert len(dist_inputs.shape) == 3, (
f'dist_inputs.shape was {dist_inputs.shape}'
f'but expected length of the tensor shape is 3')
# values of the shape float32[128,1,1]
# returns of the shape float32[128,1,1]
# and old_log_probs of the shape float32[128,1]
assert values.shape == returns.shape, (
f'values.shape was {values.shape}'
f'returns.shape was {returns.shape}')
assert returns.shape[0:2] == old_log_probs.shape, (
f'returns.shape was {returns.shape}'
f'old_log_probs.shape was {old_log_probs.shape}')
# actions is a tensor of the shape int32[128,1]
assert len(
actions.shape) == 2, f'actions.shape was {actions.shape}'
# which agrees with returns/values on the first two coordinates
assert actions.shape[0:2] == returns.shape[0:2], (
f'actions.shape was {actions.shape} and'
f'returns.shape was {returns.shape}')
ppo_objective = rl_layers.PPOObjective(
dist_inputs,
stop_gradient(values),
returns,
actions,
old_log_probs,
log_prob_fun=self._policy_dist.log_prob,
epsilon=self._epsilon,
normalize_advantages=self._normalize_advantages)
# we insist that ppo_objective is a vector of shape [128,1]
assert len(ppo_objective.shape) == 2, (
f'ppo_objective was {ppo_objective}')
# which agrees with returns/values/actions on the first two coordinates
assert ppo_objective.shape[0:2] == values.shape[0:2], (
f'ppo_objective.shape was {ppo_objective.shape} and '
f'values.shape was {values.shape}')
entropy_loss = rl_layers.EntropyLoss(
dist_inputs,
actions,
log_prob_fun=self._policy_dist.log_prob,
entropy_coeff=self._entropy_coeff,
entropy_fun=self._policy_dist.entropy)
assert jnp.ndim(
entropy_loss) == 0, f'entropy_loss was {entropy_loss}'
l2_value_loss = rl_layers.ValueLoss(
values, returns, value_loss_coeff=self._value_loss_coeff)
assert jnp.ndim(
l2_value_loss) == 0, f'l2_value_loss was {l2_value_loss}'
return -ppo_objective.mean() + l2_value_loss - entropy_loss
def AWRJointLoss(x, **unused_kwargs): # pylint: disable=invalid-name
preds, values, returns, actions, mask = x
advantages = jnp.squeeze(returns - stop_gradient(values), axis=-1)
logps = self._policy_dist.log_prob(preds, actions)
awr_loss = actor_critic.AWRLoss(beta=self._beta,
w_max=self._w_max)(
(logps, advantages,
jnp.zeros_like(logps), mask))
l2_value_loss = jnp.mean(
(returns - values)**2) * self._value_loss_coeff
return awr_loss + l2_value_loss
def f(dist_inputs, values, returns, actions, old_log_probs, mask):
"""Definition of the A2C loss."""
del old_log_probs
a2c_objective = rl_layers.A2CObjective(
dist_inputs,
stop_gradient(values),
returns,
actions,
mask,
log_prob_fun=self._policy_dist.log_prob,
normalize_advantages=self._normalize_advantages)
entropy_loss = rl_layers.EntropyLoss(
dist_inputs,
actions,
log_prob_fun=self._policy_dist.log_prob,
entropy_coeff=self._entropy_coeff,
entropy_fun=self._policy_dist.entropy)
l2_value_loss = rl_layers.ValueLoss(
values, returns, value_loss_coeff=self._value_loss_coeff)
return a2c_objective.mean() + l2_value_loss - entropy_loss
def f(log_probs, advantages, old_log_probs, mask):
if reweight: # Use new policy weights for sampled actions instead.
mask *= jnp.exp(math.stop_gradient(log_probs) - old_log_probs)
weights = jnp.minimum(awr_weights(advantages, beta), w_max)
return -jnp.sum(log_probs * weights * mask) / jnp.sum(mask)
