polish(pu):polish td3_vae config

ding/policy/td3_vae.py

@@ -269,7 +269,7 @@ class TD3VAEPolicy(DDPGPolicy):
             result['original_action'] = data['action']
             result['true_residual'] = data['next_obs'] - data['obs']
 
-            vae_loss = self._vae_model.loss_function(result, kld_weight=0.5, predict_weight=10)  # TODO(pu):weight
+            vae_loss = self._vae_model.loss_function(result, kld_weight=0.5, predict_weight=1)  # TODO(pu):weight
             # recons = args[0]
             # prediction_residual = args[1]
             # input_action = args[2]
@@ -338,7 +338,7 @@ class TD3VAEPolicy(DDPGPolicy):
                 # self.c_percentage_bound_lower = data['latent_action'].sort(dim=0)[0][int(result['recons_action'].shape[0] * 0.02), :]  # values, indices
                 # self.c_percentage_bound_upper = data['latent_action'].sort(dim=0)[0][int(result['recons_action'].shape[0] * 0.98), :]
 
-                vae_loss = self._vae_model.loss_function(result, kld_weight=0.5, predict_weight=10)  # TODO(pu):weight
+                vae_loss = self._vae_model.loss_function(result, kld_weight=0.5, predict_weight=1)  # TODO(pu):weight
 
                 loss_dict['vae_loss'] = vae_loss['loss']
                 loss_dict['reconstruction_loss'] = vae_loss['reconstruction_loss']

dizoo/box2d/lunarlander/config/lunarlander_cont_td3_vae_config.py

@@ -3,7 +3,7 @@ from ding.entry import serial_pipeline_td3_vae
 
 lunarlander_td3vae_config = dict(
     # exp_name='lunarlander_cont_ddpg_vae_lad6_wu1000_rlabelz_novaeupdatez_ns48_rvuc1000_upcr20_upcv1000_notargetnoise_collectoriginalnoise_rbs2e4_rsc',
-    exp_name='lunarlander_cont_ddpg_vae_lad6_wu1000_rlabelz_novaeupdatez_ns48_rvuc100_upcr20_upcv100_notargetnoise_collectoriginalnoise_rbs2e4_rsc',# TODO(pu) deubg
+    exp_name='lunarlander_cont_ddpg_vae_lad6_wu1000_rlabelz_novaeupdatez_ns48_rvuc100_upcr2_upcv0_notargetnoise_collectoriginalnoise_rbs2e4_rsc',# TODO(pu) deubg
     # exp_name='lunarlander_cont_ddpg_vae_lad6_wu1000_rlabelz_novaeupdatez_ns48_rvuc100_upcr2_upcv100_notargetnoise_collectoriginalnoise_rbs2e4_rsc_lsc',# TODO(pu)
 
     env=dict(
@@ -40,13 +40,13 @@ lunarlander_td3vae_config = dict(
             # rl_vae_update_circle=1,
 
             # update_per_collect_rl=50,
-            update_per_collect_rl=20,
-            # update_per_collect_rl=2,
+            # update_per_collect_rl=20,
+            update_per_collect_rl=2,
 
-            update_per_collect_vae=100,
+            # update_per_collect_vae=100,
             # update_per_collect_vae=20,
             # update_per_collect_vae=1,
-            # update_per_collect_vae=0,
+            update_per_collect_vae=0,
 
             batch_size=128,
             learning_rate_actor=1e-3,