def forward(self, inputs, tformat, loss_fn=None, hidden_states=None, **kwargs):
test_mode = kwargs["test_mode"]
avail_actions, params_aa, tformat_aa = _to_batch(inputs["avail_actions"], tformat)
x, params, tformat = _to_batch(inputs["main"], tformat)
x = F.relu(self.fc1(x))
x = self.fc2(x)
# mask policy elements corresponding to unavailable actions
n_available_actions = avail_actions.sum(dim=1, keepdim=True)
x = th.exp(x)
x = x.masked_fill(avail_actions == 0, np.sqrt(float(np.finfo(np.float32).tiny)))
x_sum = x.sum(dim=1, keepdim=True)
second_mask = (x_sum <= np.sqrt(float(np.finfo(np.float32).tiny)) * avail_actions.shape[1])
x_sum = x_sum.masked_fill(second_mask, 1.0)
x = th.div(x, x_sum)
# throw debug warning if second masking was necessary
if th.sum(second_mask.data) > 0:
if self.args.debug_verbose:
print('Warning in MACKRLNonRecurrentAgentLevel3.forward(): some sum during the softmax has been 0!')
# add softmax exploration (if switched on)
if self.args.coma_exploration_mode in ["softmax"] and not test_mode:
epsilons = inputs["epsilons"].unsqueeze(_tdim(tformat))
epsilons, _, _ = _to_batch(epsilons, tformat)
x = avail_actions * epsilons / n_available_actions + x * (1 - epsilons)
x = _from_batch(x, params, tformat)
if loss_fn is not None:
losses, _ = loss_fn(x, tformat=tformat)
return x, hidden_states, losses, tformat
def forward(self, inputs, tformat):
# _check_inputs_validity(inputs, self.input_shapes, tformat, allow_nonseq=True)
main, params, m_tformat = _to_batch(inputs.get("main"), tformat)
x = F.relu(self.fc1(main))
vvalue = self.fc2(x)
return _from_batch(vvalue, params, m_tformat), m_tformat
def forward(self, inputs, hidden_states, tformat, loss_fn=None, **kwargs):
"""
If data contains whole sequences, can pass loss_fn to forward pass in order to generate all losses
automatically.
Can either be operated in sequence mode, or operated step-by-step
"""
_check_inputs_validity(inputs, self.input_shapes, tformat)
_inputs = inputs["main"]
loss = None
t_dim = _tdim(tformat)
assert t_dim == 2, "t_dim along unsupported axis"
t_len = _inputs.shape[t_dim]
loss_x = []
output_x = []
h_list = [hidden_states]
for t in range(t_len):
x = _inputs[:, :, slice(t, t + 1), :].contiguous()
x, tformat = self.encoder({"main":x}, tformat)
x, params_x, tformat_x = _to_batch(x, tformat)
h, params_h, tformat_h = _to_batch(h_list[-1], tformat)
h = self.gru(x, h)
x = self.output(h)
h = _from_batch(h, params_h, tformat_h)
x = _from_batch(x, params_x, tformat_x)
h_list.append(h)
loss_x.append(x)
# we will not branch the variables if loss_fn is set - instead return only tensor values for x in that case
output_x.append(x) if loss_fn is None else output_x.append(x.clone())
if loss_fn is not None:
_x = th.cat(loss_x, dim=_tdim(tformat))
loss = loss_fn(_x, tformat=tformat)[0]
return th.cat(output_x, t_dim), \
th.cat(h_list[1:], t_dim), \
loss, \
tformat
def forward(self, policies, tformat):
_policies, policies_params, policies_tformat = _to_batch(
policies, tformat)
# need batch scalar product as in COMA!!!
entropy = th.bmm(
th.log(_policies).unsqueeze(1), _policies.unsqueeze(2)).squeeze(2)
ret = _from_batch(entropy, policies_params, policies_tformat)
return ret
def forward(self, inputs, tformat, **kwargs):
# _check_inputs_validity(inputs, self.input_shapes, tformat, allow_nonseq=True)
if getattr(self.args, "critic_is_recurrent", False):
_inputs = inputs.get("main")
t_dim = _tdim(tformat)
assert t_dim == 1, "t_dim along unsupported axis"
t_len = _inputs.shape[t_dim]
try:
hidden_states = kwargs["hidden_states"]
except:
pass
x_list = []
h_list = [hidden_states]
for t in range(t_len):
x = _inputs[:, slice(t, t + 1), :].contiguous()
x, params_x, tformat_x = _to_batch(x, tformat)
h, params_h, tformat_h = _to_batch(h_list[-1], tformat)
x = F.relu(self.fc1(x))
h = self.gru(x, h)
x = self.fc2(x)
h = _from_batch(h, params_h, tformat_h)
x = _from_batch(x, params_x, tformat_x)
h_list.append(h)
x_list.append(x)
return th.cat(x_list, t_dim), \
tformat
else:
main, params, m_tformat = _to_batch(inputs.get("main"), tformat)
x = F.relu(self.fc1(main))
vvalue = self.fc2(x)
return _from_batch(vvalue, params, m_tformat), m_tformat
def forward(self, inputs, tformat, loss_fn=None, hidden_states=None, **kwargs):
_check_inputs_validity(inputs, self.input_shapes, tformat)
# Execute model branch "main"
x, params, tformat = _to_batch(inputs["main"], tformat)
x = F.relu(self.fc1(x))
x = self.fc2(x)
x = _from_batch(x, params, tformat)
losses = None
if self.output_type in ["policies"]:
log_softmax = kwargs.get("log_softmax", False)
if log_softmax:
x = F.log_softmax(x, dim=_vdim(tformat))
else:
x = F.softmax(x, dim=_vdim(tformat))
if loss_fn is not None:
losses, _ = loss_fn(x, tformat=tformat)
return x, hidden_states, losses, tformat # output, hidden state, losses
def forward(self, inputs, tformat):
x, n_seq, tformat = _to_batch(inputs["main"], tformat)
x = F.relu(self.fc1(x))
return _from_batch(x, n_seq, tformat), tformat
def forward(self, inputs, actions, hidden_states, tformat, loss_fn=None, **kwargs):
seq_lens = kwargs["seq_lens"]
# TODO: How do we handle the loss propagation for recurrent layers??
# generate level 1-3 outputs
out_level1, hidden_states_level1, losses_level1, tformat_level1 = self.model_level1(inputs=inputs["level1"]["agent_input_level1"],
hidden_states=hidden_states["level1"],
loss_fn=None, #loss_fn,
tformat=tformat["level1"],
#n_agents=self.n_agents,
**kwargs)
pairwise_avail_actions = inputs["level2"]["agent_input_level2"]["avail_actions_pair"]
ttype = th.cuda.FloatTensor if pairwise_avail_actions.is_cuda else th.FloatTensor
delegation_avails = Variable(ttype(pairwise_avail_actions.shape[0],
pairwise_avail_actions.shape[1],
pairwise_avail_actions.shape[2], 1).fill_(1.0), requires_grad=False)
pairwise_avail_actions = th.cat([delegation_avails, pairwise_avail_actions], dim=_vdim(tformat["level2"]))
out_level2, hidden_states_level2, losses_level2, tformat_level2 = self.models["level2_{}".format(0)](inputs=inputs["level2"]["agent_input_level2"],
hidden_states=hidden_states["level2"],
loss_fn=None, #loss_fn,
tformat=tformat["level2"],
pairwise_avail_actions=pairwise_avail_actions,
**kwargs)
out_level3, hidden_states_level3, losses_level3, tformat_level3 = self.models["level3_{}".format(0)](inputs["level3"]["agent_input_level3"],
hidden_states=hidden_states["level3"],
loss_fn=None,
tformat=tformat["level3"],
#seq_lens=seq_lens,
**kwargs)
# for each agent pair (a,b), calculate p_a_b = p(u_a, u_b|tau_a tau_b CK_ab) = p(u_d|CK_ab)*pi(u_a|tau_a)*pi(u_b|tau_b) + p(u_ab|CK_ab)
# output dim of p_a_b is (n_agents choose 2) x bs x t x n_actions**2
# Bulk NaN masking
# out_level1[out_level1 != out_level1] = 0.0
# out_level2[out_level2 != out_level2] = 0.0
# out_level3[out_level3 != out_level3] = 0.0
# actions = actions.detach()
# actions[actions!=actions] = 0.0
p_d = out_level2[:, :, :, 0:1]
p_ab = out_level2[:, :, :, 1:-1]
_actions = actions.clone()
_actions[actions != actions] = 0.0
pi = out_level3
pi_actions_selected = out_level3.gather(_vdim(tformat_level3), _actions.long()).clone()
# pi_actions_selected[pi_actions_selected != pi_actions_selected ] = 0.0 #float("nan")
pi_actions_selected_mask = (actions != actions)
avail_actions_level3 = inputs["level3"]["agent_input_level3"]["avail_actions"].clone().data
avail_actions_selected = avail_actions_level3.gather(_vdim(tformat_level3), _actions.long()).clone()
# RENORMALIZE THIS STUFF
pi_a_cross_pi_b_list = []
pi_ab_list = []
pi_corr_list = []
# DEBUG: if prob is 0 throw a debug!
# debug
# 0 (0, 1)
# 1 (0, 2)
# 2 (1, 2)
for _i, (_a, _b) in enumerate(self.ordered_agent_pairings): #_ordered_agent_pairings(self.n_agents)[:self.args.n_pair_samples] if hasattr("n_pair_samples", self.args) else _ordered_agent_pairings(self.n_agents)): #(_ordered_agent_pairings(self.n_agents)):
# calculate pi_a_cross_pi_b # TODO: Set disallowed joint actions to NaN!
pi_a = pi[_a:_a+1].clone()
pi_b = pi[_b:_b+1].clone()
x, params_x, tformat_x = _to_batch(out_level3[_a:_a+1], tformat["level3"])
y, params_y, tformat_y = _to_batch(out_level3[_b:_b+1], tformat["level3"])
actions_masked = actions.clone()
actions_masked[actions!=actions] = 0.0
_actions_x, _actions_params, _actions_tformat = _to_batch(actions_masked[_a:_a+1].clone(), tformat["level3"])
_actions_y, _actions_params, _actions_tformat = _to_batch(actions_masked[_b:_b+1].clone(), tformat["level3"])
_x = x.gather(1, _actions_x.long())
_y = y.gather(1, _actions_y.long())
z = _x * _y
u = _from_batch(z, params_x, tformat_x)
pi_a_cross_pi_b_list.append(u)
# calculate p_ab_selected
_p_ab = p_ab[_i:_i+1].clone()
joint_actions = _action_pair_2_joint_actions((actions_masked[_a:_a+1], actions_masked[_b:_b+1]), self.n_actions)
_z = _p_ab.gather(_vdim(tformat_level2), joint_actions.long())
# Set probabilities corresponding to jointly-disallowed actions to 0.0
avail_flags = pairwise_avail_actions[_i:_i+1].gather(_vdim(tformat_level2), joint_actions.long() + 1).clone()
_z[avail_flags==0.0] = 0.0 # TODO: RENORMALIZE?
pi_ab_list.append(_z)
if not ((hasattr(self.args, "mackrl_always_delegate") and self.args.mackrl_always_delegate) or \
(hasattr(self.args, "mackrl_delegate_if_zero_ck") and self.args.mackrl_delegate_if_zero_ck)):
# Calculate corrective delegation (when unavailable actions are selected)
aa_a, _, _ = _to_batch(avail_actions_level3[_a:_a+1].clone(), tformat_level3)
aa_b, _, _ = _to_batch(avail_actions_level3[_b:_b+1].clone(), tformat_level3)
paa, params_paa, tformat_paa = _to_batch(pairwise_avail_actions[_i:_i+1].clone(), tformat_level3)
_pi_a, _, _ = _to_batch(pi_a, tformat_level3)
_pi_b, _, _ = _to_batch(pi_b, tformat_level3)
# x = th.bmm(th.unsqueeze(aa_a, 2), th.unsqueeze(aa_b, 1))
# At least one action unavailable.
diff_matrix = th.relu(paa[:, 1:-1].view(-1, self.n_actions, self.n_actions) -
th.bmm(th.unsqueeze(aa_a[:, :-1], 2), th.unsqueeze(aa_b[:, :-1], 1)))
diff_matrix = diff_matrix * _p_ab.view(-1, self.n_actions, self.n_actions)
both_unavailable = th.relu(paa[:, 1:-1].view(-1, self.n_actions, self.n_actions) -
th.add(th.unsqueeze(aa_a[:, :-1], 2), th.unsqueeze(aa_b[:, :-1], 1)))
both_unavailable = both_unavailable * _p_ab.view(-1, self.n_actions, self.n_actions)
both_unavailable_weight = th.sum(both_unavailable.view(-1, self.n_actions*self.n_actions), -1, keepdim=True)
# If neither component of the joint action is available both get resampled, with the probability of the independent actors.
correction = both_unavailable_weight * pi_actions_selected[_a:_a + 1].view(-1,1) * pi_actions_selected[_b:_b + 1].view(-1,1)
act_a = actions_masked[_a:_a + 1].clone().view(-1, 1,1).long()
act_b = actions_masked[_b:_b + 1].clone().view(-1, 1,1).long()
b_resamples = th.sum(th.gather(diff_matrix, 1, act_a.repeat([1,1, self.n_actions])),-1)
b_resamples = b_resamples * pi_actions_selected[_b:_b + 1].view(-1,1)
a_resamples = th.sum(th.gather(diff_matrix, 2, act_b.repeat([1, self.n_actions,1])), 1)
a_resamples = a_resamples * pi_actions_selected[_a:_a + 1].view(-1,1)
correction = correction + b_resamples + a_resamples
pi_corr_list.append(_from_batch(correction, params_paa, tformat_paa))
pi_a_cross_pi_b = th.cat(pi_a_cross_pi_b_list, dim=0)
pi_ab_selected = th.cat(pi_ab_list, dim=0)
p_a_b = p_d * pi_a_cross_pi_b + pi_ab_selected
agent_pairs = self.ordered_agent_pairings
if not ( (hasattr(self.args, "mackrl_always_delegate") and self.args.mackrl_always_delegate) or \
(hasattr(self.args, "mackrl_delegate_if_zero_ck") and self.args.mackrl_delegate_if_zero_ck) ):
pi_corr = th.cat(pi_corr_list, dim=0)
p_a_b += pi_corr # DEBUG
pi_c_prod_list = []
p_a_b_pair_list = []
for pairs in self.ordered_2_agent_pairings:
p_a_b_pair_list.append(reduce(operator.mul, [p_a_b[pair] for pair in pairs]))
agents_in_pairs = list(itertools.chain.from_iterable([agent_pairs[pair] for pair in pairs]))
# calculate pi_c_prod
_pi_actions_selected = pi_actions_selected.clone()
_pi_actions_selected[pi_actions_selected_mask] = 0.0
for _agent in agents_in_pairs:
_pi_actions_selected[_agent:_agent + 1] = 1.0
_k = th.prod(_pi_actions_selected, dim=_adim(tformat_level3), keepdim=True)
pi_c_prod_list.append(_k)
pi_ab_pair = th.stack(p_a_b_pair_list, dim=0)
pi_c_prod = th.cat(pi_c_prod_list, dim=0)
# next, calculate p_a_b * prod(p, -a-b)
p_prod = pi_ab_pair * pi_c_prod
# now, calculate p_a_b_c
_tmp = out_level1.transpose(_adim(tformat_level1), _vdim(tformat_level1))
_tmp[_tmp!=_tmp] = 0.0
p_a_b_c = (p_prod * _tmp).sum(dim=_adim(tformat_level1), keepdim=True)
if self.args.debug_mode in ["check_probs"]:
if not hasattr(self, "action_table"):
self.action_table = {}
if not hasattr(self, "actions_sampled"):
self.actions_sampled = 0
actions_flat = actions.view(self.n_agents, -1)
for id in range(actions_flat.shape[1]):
act = tuple(actions_flat[:, id].tolist())
if act in self.action_table:
self.action_table[act] += 1
else:
self.action_table[act] = 0
self.actions_sampled += actions_flat.shape[0]
if self.args.debug_mode in ["check_probs"]:
actions_flat = actions.view(self.n_agents, -1)
for id in range(actions_flat.shape[1]):
print("sampled: ",
self.action_table[tuple(actions_flat[:, id].tolist())] / self.actions_sampled,
" pred: ",
p_a_b_c.view(-1)[id])
# DEBUG MODE HERE!
# _check_nan(pi_c_prod)
# _check_nan(p_d)
# _check_nan(pi_ab_selected)
# _check_nan(pi_a_cross_pi_b)
# _check_nan(p_a_b)
# _check_nan(p_a_b_c)
# agent_parameters = list(self.parameters())
# pi_c_prod.sum().backward(retain_graph=True) # p_prod throws NaN
# _check_nan(agent_parameters)
# p_a_b.sum().backward(retain_graph=True)
# _check_nan(agent_parameters)
# p_prod.sum().backward(retain_graph=True)
# _check_nan(agent_parameters)
hidden_states = {"level1": hidden_states_level1,
"level2": hidden_states_level2,
"level3": hidden_states_level3
}
loss = loss_fn(policies=p_a_b_c, tformat=tformat_level3)
# loss = p_a_b_c.sum(), "a*bs*t*v"
#loss[0].sum().backward(retain_graph=True)
# loss[0].backward(retain_graph=True)
# try:
# _check_nan(agent_parameters)
# except Exception as e:
# for a, b in self.named_parameters():
# print("{}:{}".format(a, b.grad))
# a = 5
# pass
return p_a_b_c, hidden_states, loss, tformat_level3 # note: policies can have NaNs in it!!
def forward(self, inputs, hidden_states, tformat, loss_fn=None, **kwargs):
seq_lens = kwargs["seq_lens"]
try:
_check_inputs_validity(inputs, self.input_shapes, tformat)
except Exception as e:
print("Exception {} - have replaced NaNs with zeros".format(e))
for _k, _v in inputs.items():
inputs[_k][inputs[_k]!=inputs[_k]] = 0.0
pass
test_mode = kwargs["test_mode"]
_inputs = inputs["main"]
_inputs_aa = _pad(inputs["avail_actions"], tformat, seq_lens, 1.0)
loss = None
t_dim = _tdim(tformat)
assert t_dim == 2, "t_dim along unsupported axis"
t_len = _inputs.shape[t_dim]
x_list = []
h_list = [hidden_states]
for t in range(t_len):
x = _inputs[:, :, slice(t, t + 1), :].contiguous()
avail_actions = _inputs_aa[:, :, slice(t, t + 1), :].contiguous()
x, tformat = self.encoder({"main":x}, tformat)
x, params_x, tformat_x = _to_batch(x, tformat)
avail_actions, params_aa, tformat_aa = _to_batch(avail_actions, tformat)
h, params_h, tformat_h = _to_batch(h_list[-1], tformat)
h = self.gru(x, h)
x = self.output(h)
# # mask policy elements corresponding to unavailable actions
# n_available_actions = avail_actions.sum(dim=1, keepdim=True)
# x = th.exp(x)
# x = x.masked_fill(avail_actions == 0, np.sqrt(float(np.finfo(np.float32).tiny)))
# x_sum = x.sum(dim=1, keepdim=True)
# second_mask = (x_sum <= np.sqrt(float(np.finfo(np.float32).tiny))*avail_actions.shape[1])
# x_sum = x_sum.masked_fill(second_mask, 1.0)
# x = th.div(x, x_sum)
n_available_actions = avail_actions.sum(dim=1, keepdim=True)
x = x - (1 - avail_actions) * 1e30
x = F.softmax(x, 1)
# throw debug warning if second masking was necessary
# if th.sum(second_mask.data) > 0:
# if self.args.debug_verbose:
# print('Warning in MACKRLRecurrentAgentLevel3.forward(): some sum during the softmax has been 0!')
# add softmax exploration (if switched on)
if self.args.mackrl_exploration_mode_level3 in ["softmax"] and not test_mode:
epsilons = inputs["epsilons_central_level3"].unsqueeze(_tdim(tformat)).detach()
epsilons, _, _ = _to_batch(epsilons, tformat)
# n_available_actions[n_available_actions==0.0] = 1 #np.sqrt(float(np.finfo(np.float32).tiny))
#if th.sum(th.sum(n_available_actions, dim=-1) > 0):
# a = 6
# pass
#if th.sum(n_available_actions == 0.0) > 0:
# a = 5
# pass
x = avail_actions.detach() * epsilons / n_available_actions + x * (1 - epsilons) # avail_actions * epsilons / n_available_actions + x * (1 - epsilons)
h = _from_batch(h, params_h, tformat_h)
x = _from_batch(x, params_x, tformat_x)
h_list.append(h)
x_list.append(x)
if loss_fn is not None:
_x = th.cat(x_list, dim=_tdim(tformat))
loss = loss_fn(_x, tformat=tformat)[0]
return th.cat(x_list, t_dim), \
th.cat(h_list[1:], t_dim), \
loss, \
tformat
def forward(self, inputs, hidden_states, tformat, loss_fn=None, **kwargs):
seq_lens = kwargs["seq_lens"]
try:
_check_inputs_validity(inputs, self.input_shapes, tformat)
except Exception as e:
print("Exception {} - have replaced NaNs with zeros".format(e))
for _k, _v in inputs.items():
inputs[_k][inputs[_k]!=inputs[_k]] = 0.0
pass
test_mode = kwargs["test_mode"]
pairwise_avail_actions = _pad(kwargs["pairwise_avail_actions"].detach(), tformat, seq_lens, 1.0)
pairwise_avail_actions.requires_grad = False
_inputs = inputs["main"]
loss = None
t_dim = _tdim(tformat)
assert t_dim == 2, "t_dim along unsupported axis"
t_len = _inputs.shape[t_dim]
x_list = []
h_list = [hidden_states]
for t in range(t_len):
x = _inputs[:, :, slice(t, t + 1), :].contiguous()
avail_actions = pairwise_avail_actions[:, :, slice(t, t + 1), :].contiguous().detach()
x, tformat = self.encoder({"main":x}, tformat)
x, params_x, tformat_x = _to_batch(x, tformat)
avail_actions, params_aa, tformat_aa = _to_batch(avail_actions, tformat)
h, params_h, tformat_h = _to_batch(h_list[-1], tformat)
h = self.gru(x, h)
x = self.output(h)
if getattr(self.args, "mackrl_logit_bias", 0.0) != 0.0:
x = th.cat([x[:, 0:1] + self.args.mackrl_logit_bias, x[:, 1:]], dim=1)
n_available_actions = avail_actions.sum(dim=1, keepdim=True)
x = x - (1 - avail_actions) * 1e30
x = F.softmax(x, 1)
# add softmax exploration (if switched on)
if self.args.mackrl_exploration_mode_level2 in ["softmax"] and not test_mode:
epsilons = inputs["epsilons_central_level2"].unsqueeze(_tdim(tformat)).detach()
epsilons, _, _ = _to_batch(epsilons, tformat)
n_available_actions[n_available_actions==0.0] = 1.0
x = avail_actions.detach() * epsilons / n_available_actions + x * (1 - epsilons)
h = _from_batch(h, params_h, tformat_h)
x = _from_batch(x, params_x, tformat_x)
h_list.append(h)
x_list.append(x)
x_cat = th.cat(x_list, t_dim)
if hasattr(self.args, "mackrl_always_delegate") and self.args.mackrl_always_delegate:
x_cat[:, :, :, 0] = 1.0
x_cat[:, :, :, 1:] = 0.0
if loss_fn is not None:
loss = loss_fn(x_cat, tformat=tformat)[0]
return x_cat, \
th.cat(h_list[1:], t_dim), \
loss, \
tformat
def forward(self, inputs, hidden_states, tformat, loss_fn=None, **kwargs):
try:
_check_inputs_validity(inputs, self.input_shapes, tformat)
except Exception as e:
print("Exception {} - have replaced NaNs with zeros".format(e))
for _k, _v in inputs.items():
inputs[_k][inputs[_k]!=inputs[_k]] = 0.0
pass
test_mode = kwargs["test_mode"]
n_agents = kwargs["n_agents"]
if len(inputs["main"].shape) == 3:
_inputs = inputs["main"].unsqueeze(0) # as agent dimension is lacking
else:
_inputs = inputs["main"]
#_inputs_aa = inputs["avail_actions"]
loss = None
t_dim = _tdim(tformat)
assert t_dim == 2, "t_dim along unsupported axis"
t_len = _inputs.shape[t_dim]
x_list = []
h_list = [hidden_states]
for t in range(t_len):
x = _inputs[:, :, slice(t, t + 1), :].contiguous()
#avail_actions = _inputs_aa[:, :, slice(t, t + 1), :].contiguous()
x, tformat = self.encoder({"main":x}, tformat)
x, params_x, tformat_x = _to_batch(x, tformat)
#avail_actions, params_aa, tformat_aa = _to_batch(avail_actions, tformat)
h, params_h, tformat_h = _to_batch(h_list[-1], tformat)
h = self.gru(x, h)
x = self.output(h)
# mask policy elements corresponding to unavailable actions
#n_available_actions = avail_actions.detach().sum(dim=1, keepdim=True)
# DEBUG
x = F.softmax(x, 1)
# Alternative variant
#x = th.nn.functional.softmax(x).clone()
#x.masked_fill_(avail_actions.long() == 0, float(np.finfo(np.float32).tiny))
#x = th.div(x, x.sum(dim=1, keepdim=True))
if self.args.mackrl_exploration_mode_level1 in ["softmax"] and not test_mode:
epsilons = inputs["epsilons_central_level1"].unsqueeze(_tdim("bs*t*v")).detach()
epsilons, _, _ = _to_batch(epsilons, "bs*t*v")
x = epsilons / x.shape[-1] + x * (1 - epsilons)
if hasattr(self.args, "mackrl_fix_level1_pair") and self.args.mackrl_fix_level1_pair:
#x.fill_(0.0)
x[:,1:] = 0.0
x[:, 0] = 1.0
h = _from_batch(h, params_h, tformat_h)
x = _from_batch(x, params_x, tformat_x)
h_list.append(h)
x_list.append(x)
if loss_fn is not None:
_x = th.cat(x_list, dim=_tdim(tformat))
loss = loss_fn(_x, tformat=tformat)[0]
return th.cat(x_list, t_dim), \
th.cat(h_list[1:], t_dim), \
loss, \
tformat