Improve const-correctness

Omegastick · Omegastick · commit 37ce31069e7c · 2019-10-29T16:04:22.000+09:00
diff --git a/include/cpprl/model/policy.h b/include/cpprl/model/policy.h
@@ -33,17 +33,17 @@ class PolicyImpl : public nn::Module
 
     std::vector<torch::Tensor> act(torch::Tensor inputs,
                                    torch::Tensor rnn_hxs,
-                                   torch::Tensor masks);
+                                   torch::Tensor masks) const;
     std::vector<torch::Tensor> evaluate_actions(torch::Tensor inputs,
                                                 torch::Tensor rnn_hxs,
                                                 torch::Tensor masks,
-                                                torch::Tensor actions);
+                                                torch::Tensor actions) const;
     torch::Tensor get_probs(torch::Tensor inputs,
                             torch::Tensor rnn_hxs,
-                            torch::Tensor masks);
+                            torch::Tensor masks) const;
     torch::Tensor get_values(torch::Tensor inputs,
                              torch::Tensor rnn_hxs,
-                             torch::Tensor masks);
+                             torch::Tensor masks) const;
     void update_observation_normalizer(torch::Tensor observations);
 
     inline bool is_recurrent() const { return base->is_recurrent(); }
diff --git a/include/cpprl/observation_normalizer.h b/include/cpprl/observation_normalizer.h
@@ -23,7 +23,7 @@ class ObservationNormalizerImpl : public torch::nn::Module
                               float clip = 10.);
     explicit ObservationNormalizerImpl(const std::vector<ObservationNormalizer> &others);
 
-    torch::Tensor process_observation(torch::Tensor observation);
+    torch::Tensor process_observation(torch::Tensor observation) const;
     std::vector<float> get_mean() const;
     std::vector<float> get_variance() const;
     void update(torch::Tensor observations);
diff --git a/src/model/policy.cpp b/src/model/policy.cpp
@@ -51,7 +51,7 @@ PolicyImpl::PolicyImpl(ActionSpace action_space,
 
 std::vector<torch::Tensor> PolicyImpl::act(torch::Tensor inputs,
                                            torch::Tensor rnn_hxs,
-                                           torch::Tensor masks)
+                                           torch::Tensor masks) const
 {
     if (observation_normalizer)
     {
@@ -83,7 +83,7 @@ std::vector<torch::Tensor> PolicyImpl::act(torch::Tensor inputs,
 std::vector<torch::Tensor> PolicyImpl::evaluate_actions(torch::Tensor inputs,
                                                         torch::Tensor rnn_hxs,
                                                         torch::Tensor masks,
-                                                        torch::Tensor actions)
+                                                        torch::Tensor actions) const
 {
     if (observation_normalizer)
     {
@@ -116,7 +116,7 @@ std::vector<torch::Tensor> PolicyImpl::evaluate_actions(torch::Tensor inputs,
 
 torch::Tensor PolicyImpl::get_probs(torch::Tensor inputs,
                                     torch::Tensor rnn_hxs,
-                                    torch::Tensor masks)
+                                    torch::Tensor masks) const
 {
     if (observation_normalizer)
     {
@@ -131,7 +131,7 @@ torch::Tensor PolicyImpl::get_probs(torch::Tensor inputs,
 
 torch::Tensor PolicyImpl::get_values(torch::Tensor inputs,
                                      torch::Tensor rnn_hxs,
-                                     torch::Tensor masks)
+                                     torch::Tensor masks) const
 {
     if (observation_normalizer)
     {
diff --git a/src/observation_normalizer.cpp b/src/observation_normalizer.cpp
@@ -66,7 +66,7 @@ ObservationNormalizerImpl::ObservationNormalizerImpl(const std::vector<Observati
     rms->set_count(total_count);
 }
 
-torch::Tensor ObservationNormalizerImpl::process_observation(torch::Tensor observation)
+torch::Tensor ObservationNormalizerImpl::process_observation(torch::Tensor observation) const
 {
     auto normalized_obs = (observation - rms->get_mean()) /
                           torch::sqrt(rms->get_variance() + 1e-8);

Original file line number	Diff line number	Diff line change
`@@ -51,7 +51,7 @@ PolicyImpl::PolicyImpl(ActionSpace action_space,`
`51`	`51`
`52`	`52`	`std::vector<torch::Tensor> PolicyImpl::act(torch::Tensor inputs,`
`53`	`53`	`torch::Tensor rnn_hxs,`
`54`		`- torch::Tensor masks)`
	`54`	`+ torch::Tensor masks) const`
`55`	`55`	`{`
`56`	`56`	`if (observation_normalizer)`
`57`	`57`	`{`
`@@ -83,7 +83,7 @@ std::vector<torch::Tensor> PolicyImpl::act(torch::Tensor inputs,`
`83`	`83`	`std::vector<torch::Tensor> PolicyImpl::evaluate_actions(torch::Tensor inputs,`
`84`	`84`	`torch::Tensor rnn_hxs,`
`85`	`85`	`torch::Tensor masks,`
`86`		`- torch::Tensor actions)`
	`86`	`+ torch::Tensor actions) const`
`87`	`87`	`{`
`88`	`88`	`if (observation_normalizer)`
`89`	`89`	`{`
`@@ -116,7 +116,7 @@ std::vector<torch::Tensor> PolicyImpl::evaluate_actions(torch::Tensor inputs,`
`116`	`116`
`117`	`117`	`torch::Tensor PolicyImpl::get_probs(torch::Tensor inputs,`
`118`	`118`	`torch::Tensor rnn_hxs,`
`119`		`- torch::Tensor masks)`
	`119`	`+ torch::Tensor masks) const`
`120`	`120`	`{`
`121`	`121`	`if (observation_normalizer)`
`122`	`122`	`{`
`@@ -131,7 +131,7 @@ torch::Tensor PolicyImpl::get_probs(torch::Tensor inputs,`
`131`	`131`
`132`	`132`	`torch::Tensor PolicyImpl::get_values(torch::Tensor inputs,`
`133`	`133`	`torch::Tensor rnn_hxs,`
`134`		`- torch::Tensor masks)`
	`134`	`+ torch::Tensor masks) const`
`135`	`135`	`{`
`136`	`136`	`if (observation_normalizer)`
`137`	`137`	`{`
Original file line number	Diff line number	Diff line change
`@@ -66,7 +66,7 @@ ObservationNormalizerImpl::ObservationNormalizerImpl(const std::vector<Observati`
`66`	`66`	`rms->set_count(total_count);`
`67`	`67`	`}`
`68`	`68`
`69`		`-torch::Tensor ObservationNormalizerImpl::process_observation(torch::Tensor observation)`
	`69`	`+torch::Tensor ObservationNormalizerImpl::process_observation(torch::Tensor observation) const`
`70`	`70`	`{`
`71`	`71`	`auto normalized_obs = (observation - rms->get_mean()) /`
`72`	`72`	`torch::sqrt(rms->get_variance() + 1e-8);`