Tidy up yaml configs (#121)

examples/3_train_policy.py

@@ -38,15 +38,13 @@ policy = DiffusionPolicy(cfg, lr_scheduler_num_training_steps=training_steps, da
 policy.train()
 policy.to(device)
 
-optimizer = torch.optim.Adam(
-    policy.diffusion.parameters(), cfg.lr, cfg.adam_betas, cfg.adam_eps, cfg.adam_weight_decay
-)
+optimizer = torch.optim.Adam(policy.parameters(), lr=1e-4)
 
 # Create dataloader for offline training.
 dataloader = torch.utils.data.DataLoader(
     dataset,
     num_workers=4,
-    batch_size=cfg.batch_size,
+    batch_size=64,
     shuffle=True,
     pin_memory=device != torch.device("cpu"),
     drop_last=True,