Views
No views yet
| Data Setting | Episodes per Task | Total Episodes |
|---|---|---|
| Clean Setting | 50 | 2,500 |
| Randomized Setting | 500 | 25,000 |
| Total | 550 | 27,500 |
TrainConfig(
name="pi05_base_finetune_on_robotwin_clean_randomized_joint_training",
project_name="pi05_finetune",
exp_name="robotwin_clean_randomized_joint_training",
model=pi0_config.Pi0Config(
pi05=True,
action_horizon=32,
),
weight_loader=weight_loaders.CheckpointWeightLoader("./pi05_base/params"),
pytorch_weight_path="./pi05_base_torch",
lr_schedule=_optimizer.ConstantScheduleWithWarmup, # we defined this scheduler ourselves
#optimizer=_optimizer.AdamW,
data=LeRobotAlohaDataConfig(
repo_id= "clean_randomized_joint_training",
data_dir= HF_LEROBOT_HOME / "robotwin" ,
multi_task=True, # we use the MultiLeRobotDataset and therefore modified the official repo slightly to better start the train
base_config=DataConfig(
prompt_from_task=True,
random_prompt_from_task=True, # we add this option for the training setting of RoboTwin(every episode of the same task randomly selects instruction for training from the same bunch of instructions)
),
assets=AssetsConfig(
assets_dir="./assets/pi05_base_finetune_on_robotwin_clean_randomized_joint_training",
asset_id="robotwin_clean_randomized_joint_training",
),
adapt_to_pi=True,
default_prompt=DEFAULT_PROMPT,
use_delta_joint_actions=True,
repack_transforms=_transforms.Group(
inputs=[
_transforms.RepackTransform(
{
"images": {
"cam_high": "high_image",
"cam_left_wrist": "left_wrist_image",
"cam_right_wrist": "right_wrist_image",
},
"state": "state",
"actions": "actions",
"prompt": "prompt",
}
),
]
),
action_sequence_keys=("actions",)
),
seed=42,
batch_size=128,
num_workers=16,
num_train_steps=1000000,
log_interval=100,
val_interval=1000,
save_interval=5000,
keep_period=5000,
resume=True,
wandb_enabled=True,
),