simplify pre-training dataset, use npz

cfg/gym/pretrain/halfcheetah-medium-v2/pre_diffusion_mlp.yaml

@@ -7,7 +7,7 @@ _target_: agent.pretrain.train_diffusion_agent.TrainDiffusionAgent
 
 name: ${env}_pre_diffusion_mlp_ta${horizon_steps}_td${denoising_steps}
 logdir: ${oc.env:DPPO_LOG_DIR}/gym-pretrain/${name}/${now:%Y-%m-%d}_${now:%H-%M-%S}_${seed}
-train_dataset_path: ${oc.env:DPPO_DATA_DIR}/gym/${env}/train.pkl
+train_dataset_path: ${oc.env:DPPO_DATA_DIR}/gym/${env}/train.npz
 
 seed: 42
 device: cuda:0
@@ -64,7 +64,7 @@ ema:
   decay: 0.995
 
 train_dataset:
-  _target_: agent.dataset.sequence.StitchedActionSequenceDataset
+  _target_: agent.dataset.sequence.StitchedSequenceDataset
   dataset_path: ${train_dataset_path}
   horizon_steps: ${horizon_steps}
   cond_steps: ${cond_steps}