prepare exp simulation from delay, input for offline data comes from …

…recorded data
lasgroup · Oct 18, 2023 · 3c6794d · 3c6794d
1 parent 8ac4d24
commit 3c6794d
Show file tree

Hide file tree

Showing 2 changed files with 12 additions and 6 deletions.
diff --git a/experiments/offline_rl_from_recorded_data/exp.py b/experiments/offline_rl_from_recorded_data/exp.py
@@ -44,6 +44,7 @@ def experiment(horizon_len: int,
                num_epochs: int = 50,
                max_train_steps: int = 100_000,
                length_scale_aditive_sim_gp: float = 1.0,
+               input_from_recorded_data: int = 1,
                ):
     bnn_train_steps = min(num_epochs * num_offline_collected_transitions, max_train_steps)
 
@@ -119,6 +120,7 @@ def experiment(horizon_len: int,
                        num_epochs=num_epochs,
                        max_train_steps=max_train_steps,
                        length_scale_aditive_sim_gp=length_scale_aditive_sim_gp,
+                       input_from_recorded_data=input_from_recorded_data
                        )
 
     total_config = SAC_KWARGS | config_dict
@@ -135,8 +137,9 @@ def experiment(horizon_len: int,
     key, key_data_seed = jr.split(key, 2)
     int_data_seed = jr.randint(key_data_seed, (), minval=0, maxval=2 ** 13 - 1)
     if data_from_simulation:
+        source = 'racecar_from_true_input_data' if input_from_recorded_data else 'racecar_actionstack'
         x_train, y_train, x_test, y_test, sim = provide_data_and_sim(
-            data_source='racecar_from_true_input_data',
+            data_source=source,
             data_spec={'num_samples_train': num_offline_collected_transitions,
                        'use_hf_sim': bool(high_fidelity),
                        'sampling': 'iid',
@@ -281,7 +284,8 @@ def main(args):
         bandwidth_svgd=args.bandwidth_svgd,
         num_epochs=args.num_epochs,
         max_train_steps=args.max_train_steps,
-        length_scale_aditive_sim_gp=args.length_scale_aditive_sim_gp
+        length_scale_aditive_sim_gp=args.length_scale_aditive_sim_gp,
+        input_from_recorded_data=args.input_from_recorded_data,
     )
 
 
@@ -317,5 +321,6 @@ def main(args):
     parser.add_argument('--num_epochs', type=int, default=20)
     parser.add_argument('--max_train_steps', type=int, default=2_000)
     parser.add_argument('--length_scale_aditive_sim_gp', type=float, default=1.0)
+    parser.add_argument('--input_from_recorded_data', type=int, default=1)
     args = parser.parse_args()
     main(args)
diff --git a/experiments/offline_rl_from_recorded_data/launcher.py b/experiments/offline_rl_from_recorded_data/launcher.py
@@ -1,7 +1,7 @@
 import exp
 from experiments.util import generate_run_commands, generate_base_command, dict_permutations
 
-PROJECT_NAME = 'OfflineRLCarV2BigExp'
+PROJECT_NAME = 'OfflineRLSimulationWithDelay'
 
 _applicable_configs = {
     'horizon_len': [200],
@@ -17,18 +17,19 @@
     'margin_factor': [20.0],
     'ctrl_cost_weight': [0.005],
     'ctrl_diff_weight': [1.0],
-    'num_offline_collected_transitions': [200, 400, 800, 1600, 2_000, 2_500, 3_000, 4_000, 5_000, 7_500, 10_000],
+    'num_offline_collected_transitions': [20_000],
     'test_data_ratio': [0.0],
     'eval_on_all_offline_data': [1],
     'eval_only_on_init_states': [1],
     'share_of_x0s_in_sac_buffer': [0.5],
     'bnn_batch_size': [32],
     'likelihood_exponent': [1.0],
     'train_sac_only_from_init_states': [0],
-    'data_from_simulation': [0],
+    'data_from_simulation': [1],
     'num_frame_stack': [3],
     'bandwidth_svgd': [0.05, 0.1, 0.2],
-    'length_scale_aditive_sim_gp': [0.2, 1.0, 5.0]
+    'length_scale_aditive_sim_gp': [0.2, 1.0, 5.0],
+    'input_from_recorded_data': [1],
 }
 
 _applicable_configs_no_sim_prior = {'use_sim_prior': [0],