Fixed parametric dqn cartpole config

Summary: Use parameters of discrete_dqn_cartpole_v0.json with model_type set to pytorch_parametric_dqn

Reviewed By: badrinarayan, kittipatv

Differential Revision: D20089999

fbshipit-source-id: 43b7f08c552535e5f96847600beafc8b77b65a5a

ml/rl/test/gym/parametric_dqn_cartpole_v0.json

@@ -1,12 +1,11 @@
 {
   "env": "CartPole-v0",
   "model_type": "pytorch_parametric_dqn",
-  "max_replay_memory_size": 10000,
-  "use_gpu": true,
+  "max_replay_memory_size": 20000,
+  "use_gpu": false,
   "rl": {
     "gamma": 0.99,
     "target_update_rate": 0.1,
-
     "maxq_learning": true,
     "epsilon": 0.05,
     "temperature": 0.35,
@@ -17,30 +16,22 @@
     "dueling_architecture": false
   },
   "training": {
-    "layers": [
-      -1,
-      128,
-      64,
-      -1
-    ],
-    "activations": [
-      "relu",
-      "relu",
-      "linear"
-    ],
-    "minibatch_size": 1024,
-    "learning_rate": 0.001,
+    "layers": [-1, 128, 64, -1],
+    "activations": ["relu", "relu", "linear"],
+    "minibatch_size": 512,
+    "learning_rate": 0.01,
     "optimizer": "ADAM",
     "lr_decay": 0.999
   },
   "run_details": {
     "num_episodes": 200,
     "max_steps": 200,
-    "train_every_ts": 1,
+    "train_every_ts": 3,
     "train_after_ts": 1,
-    "test_every_ts": 2000,
-    "test_after_ts": 1,
+    "test_every_ts": 400,
+    "test_after_ts": 1000,
     "num_train_batches": 1,
-    "avg_over_num_episodes": 100
+    "avg_over_num_episodes": 25,
+    "offline_train_epochs": 7
   }
 }