Delete 2026.04.05

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/.hydra/config.yaml

@@ -1,168 +0,0 @@
-name: train_bdpo_unet_lowdim
-_target_: llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace
-obs_dim: ${task.obs_dim}
-action_dim: ${task.action_dim}
-task_name: ${task.name}
-exp_name: default
-horizon: 1
-n_obs_steps: 1
-n_action_steps: 1
-n_latency_steps: 0
-past_action_visible: false
-policy:
-  _target_: llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy
-  model:
-    _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
-    input_dim: ${task.action_dim}
-    local_cond_dim: null
-    global_cond_dim: '${eval: ${task.obs_dim}*${n_obs_steps}}'
-    diffusion_step_embed_dim: 256
-    down_dims:
-    - 256
-    - 512
-    - 1024
-    kernel_size: 5
-    n_groups: 8
-    cond_predict_scale: true
-  noise_scheduler:
-    _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
-    num_train_timesteps: 100
-    beta_start: 0.0001
-    beta_end: 0.02
-    beta_schedule: squaredcos_cap_v2
-    variance_type: fixed_small
-    clip_sample: true
-    prediction_type: epsilon
-  obs_dim: ${obs_dim}
-  action_dim: ${action_dim}
-  horizon: ${horizon}
-  n_obs_steps: ${n_obs_steps}
-  n_action_steps: ${n_action_steps}
-  hidden_size:
-  - 256
-  - 256
-  activation: relu
-  gamma: ${task.env_runner.discount}
-  tau: 0.005
-  eta: 1.0
-  rho: 1.0
-  target_reduction: min
-  max_q_backup: false
-  actor_update_interval: 1
-  value_warmup_steps: 1000
-  num_eval_sample_steps: 100
-  act_limit: null
-  squash_action: false
-  init_weight: 0.01
-  checkpoint_path: null
-  behavior_checkpoint_path: data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0055-val_loss=0.215.ckpt
-  init_actor_from_behavior: true
-dataloader:
-  batch_size: 256
-  num_workers: 0
-  shuffle: true
-  pin_memory: false
-  persistent_workers: false
-val_dataloader:
-  batch_size: 256
-  num_workers: 0
-  shuffle: true
-  pin_memory: false
-  persistent_workers: false
-optimizer:
-  _target_: torch.optim.Adam
-  lr: 0.0001
-  betas:
-  - 0.9
-  - 0.999
-  eps: 1.0e-08
-  weight_decay: 0.0
-critic_optimizer:
-  _target_: torch.optim.Adam
-  lr: 0.0001
-  betas:
-  - 0.9
-  - 0.999
-  eps: 1.0e-08
-  weight_decay: 0.0
-value_optimizer:
-  _target_: torch.optim.Adam
-  lr: 0.0001
-  betas:
-  - 0.9
-  - 0.999
-  eps: 1.0e-08
-  weight_decay: 0.0
-training:
-  device: cuda:0
-  seed: 45
-  debug: false
-  resume: false
-  lr_scheduler: constant
-  lr_warmup_steps: 500
-  num_epochs: 201
-  gradient_accumulate_every: 1
-  grad_norm_clip: 5.0
-  rollout_every: 10
-  checkpoint_every: 10
-  val_every: 1
-  sample_every: 10
-  sample_max_batch: 128
-  max_train_steps: null
-  max_val_steps: null
-  tqdm_interval_sec: 1.0
-logging:
-  project: ${task.name}-bdpo-training
-  resume: true
-  mode: online
-  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
-  tags:
-  - ${name}
-  - ${task_name}
-  - ${exp_name}
-  id: null
-  group: null
-checkpoint:
-  topk:
-    monitor_key: test_success_rate
-    mode: max
-    k: 5
-    format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
-  save_last_ckpt: true
-  save_last_snapshot: false
-multi_run:
-  run_dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
-  wandb_name_base: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
-task:
-  name: adroit-hand-hammer-v1
-  obs_dim: 46
-  action_dim: 26
-  env_runner:
-    _target_: llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner
-    env_name: llf-adroit-adroit-hand-hammer-v1
-    n_train: 10
-    n_test: 50
-    n_envs: 10
-    max_steps: 150
-    n_obs_steps: ${n_obs_steps}
-    n_action_steps: ${n_action_steps}
-    instruction_type: b
-    feedback_type:
-    - hp
-    - hn
-    - fp
-    visual: false
-    discount: 0.99
-  dataset:
-    _target_: llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset
-    data_path: datasets/adroit-hand-hammer-v1-general-reward-trans.pt
-    data_path2: datasets/adroit-hand-hammer-v1-reward-trans.pt
-    horizon: ${horizon}
-    pad_before: ${eval:'${n_obs_steps}-1'}
-    pad_after: ${eval:'${n_action_steps}-1'}
-    obs_eef_target: true
-    use_manual_normalizer: false
-    val_ratio: 0.1
-    dummy_normalizer: false
-  instructor:
-    _target_: llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/.hydra/hydra.yaml

@@ -1,156 +0,0 @@
-hydra:
-  run:
-    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
-  sweep:
-    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
-    subdir: ${hydra.job.num}
-  launcher:
-    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
-  sweeper:
-    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
-    max_batch_size: null
-    params: null
-  help:
-    app_name: ${hydra.job.name}
-    header: '${hydra.help.app_name} is powered by Hydra.
-
-      '
-    footer: 'Powered by Hydra (https://hydra.cc)
-
-      Use --hydra-help to view Hydra specific help
-
-      '
-    template: '${hydra.help.header}
-
-      == Configuration groups ==
-
-      Compose your configuration from those groups (group=option)
-
-
-      $APP_CONFIG_GROUPS
-
-
-      == Config ==
-
-      Override anything in the config (foo.bar=value)
-
-
-      $CONFIG
-
-
-      ${hydra.help.footer}
-
-      '
-  hydra_help:
-    template: 'Hydra (${hydra.runtime.version})
-
-      See https://hydra.cc for more info.
-
-
-      == Flags ==
-
-      $FLAGS_HELP
-
-
-      == Configuration groups ==
-
-      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
-      to command line)
-
-
-      $HYDRA_CONFIG_GROUPS
-
-
-      Use ''--cfg hydra'' to Show the Hydra config.
-
-      '
-    hydra_help: ???
-  hydra_logging:
-    version: 1
-    formatters:
-      simple:
-        format: '[%(asctime)s][HYDRA] %(message)s'
-    handlers:
-      console:
-        class: logging.StreamHandler
-        formatter: simple
-        stream: ext://sys.stdout
-    root:
-      level: INFO
-      handlers:
-      - console
-    loggers:
-      logging_example:
-        level: DEBUG
-    disable_existing_loggers: false
-  job_logging:
-    version: 1
-    formatters:
-      simple:
-        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
-    handlers:
-      console:
-        class: logging.StreamHandler
-        formatter: simple
-        stream: ext://sys.stdout
-      file:
-        class: logging.FileHandler
-        formatter: simple
-        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
-    root:
-      level: INFO
-      handlers:
-      - console
-      - file
-    disable_existing_loggers: false
-  env: {}
-  mode: RUN
-  searchpath: []
-  callbacks: {}
-  output_subdir: .hydra
-  overrides:
-    hydra:
-    - hydra.mode=RUN
-    task:
-    - training.seed=45
-    - policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\=0055-val_loss\=0.215.ckpt
-  job:
-    name: train
-    chdir: null
-    override_dirname: policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\=0055-val_loss\=0.215.ckpt,training.seed=45
-    id: ???
-    num: ???
-    config_name: bdpo_adroit-hand-hammer-v1_gen.yaml
-    env_set: {}
-    env_copy: []
-    config:
-      override_dirname:
-        kv_sep: '='
-        item_sep: ','
-        exclude_keys: []
-  runtime:
-    version: 1.2.0
-    version_base: '1.2'
-    cwd: /tmp2/chyang/workspace/LLM-BC
-    config_sources:
-    - path: hydra.conf
-      schema: pkg
-      provider: hydra
-    - path: /tmp2/chyang/workspace/LLM-BC/config/main_table
-      schema: file
-      provider: main
-    - path: ''
-      schema: structured
-      provider: schema
-    output_dir: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-    choices:
-      hydra/env: default
-      hydra/callbacks: null
-      hydra/job_logging: default
-      hydra/hydra_logging: default
-      hydra/hydra_help: default
-      hydra/help: default
-      hydra/sweeper: basic
-      hydra/launcher: basic
-      hydra/output: default
-  verbose: false

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/.hydra/overrides.yaml

@@ -1,2 +0,0 @@
-- training.seed=45
-- policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\=0055-val_loss\=0.215.ckpt

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0000-test_success_rate=0.300.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ab983ad24178337d87c0786bb0806f9c1e503ce9a96576fb997bda65bd9667a9
-size 1055778720

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0010-test_success_rate=0.260.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:148ceda9d6dff960ca2ccff4be27c56d5dea10412bde3cf8fe4aa0359a6dc3f4
-size 1055778720

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0020-test_success_rate=0.300.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:acb55fb904569722c484398ec42b8417508de98906fc26a333cef9597ced2b5c
-size 1580582693

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0030-test_success_rate=0.280.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f30cc47b5a51be6dff657cf4e9e5e4a92082bc6ab97bab90d998a05f6d5b350c
-size 1580582693

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0040-test_success_rate=0.240.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:aa91a0661f2df1acaf2739892f21c5286c28e660bf6d8e20b3faa3865b64cb98
-size 1580582693

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/latest.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f8ffdd359d8d361b7fafd12f3c0ede60a66c242b6b00122391dbf09fcee8119b
-size 1580582693

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/train.log

@@ -1,11 +0,0 @@
-[2026-04-05 19:47:22,077][numexpr.utils][INFO] - Note: NumExpr detected 20 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
-[2026-04-05 19:47:22,077][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
-[2026-04-05 19:47:22,750][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.558467e+07
-[2026-04-05 19:47:28,683][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.558467e+07
-[2026-04-05 19:47:32,007][OpenGL.acceleratesupport][WARNING] - Incompatible version of OpenGL_accelerate found, need at least (3, 1, 10) found (3, 1, 9)
-[2026-04-05 19:47:32,007][OpenGL.acceleratesupport][INFO] - No OpenGL_accelerate module loaded: Old version of OpenGL_accelerate
-[2026-04-05 19:47:36,329][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
-[2026-04-05 19:47:36,339][absl][INFO] - MuJoCo library version is: 2.3.7
-[2026-04-05 19:47:36,507][datasets][INFO] - PyTorch version 2.2.2 available.
-[2026-04-05 19:47:36,508][datasets][INFO] - TensorFlow version 2.15.1 available.
-[2026-04-05 19:47:36,509][datasets][INFO] - JAX version 0.4.30 available.

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/debug-internal.log

@@ -1,16 +0,0 @@
-{"time":"2026-04-05T19:47:37.711575744+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
-{"time":"2026-04-05T19:47:37.711598693+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug-core.log"}
-{"time":"2026-04-05T19:47:37.820659395+08:00","level":"INFO","msg":"created new stream","id":"4ns1dtmm"}
-{"time":"2026-04-05T19:47:37.820714248+08:00","level":"INFO","msg":"stream: started","id":"4ns1dtmm"}
-{"time":"2026-04-05T19:47:37.820739406+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-05T19:47:37.820815969+08:00","level":"INFO","msg":"sender: started","stream_id":"4ns1dtmm"}
-{"time":"2026-04-05T19:47:37.820853585+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-05T19:47:38.804274688+08:00","level":"INFO","msg":"Starting system monitor"}
-{"time":"2026-04-06T14:06:45.702933515+08:00","level":"INFO","msg":"stream: closing","id":"4ns1dtmm"}
-{"time":"2026-04-06T14:06:45.702982796+08:00","level":"INFO","msg":"Stopping system monitor"}
-{"time":"2026-04-06T14:06:45.703854988+08:00","level":"INFO","msg":"Stopped system monitor"}
-{"time":"2026-04-06T14:06:49.250748912+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
-{"time":"2026-04-06T14:06:49.546933574+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-06T14:06:49.546991191+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-06T14:06:49.547007286+08:00","level":"INFO","msg":"sender: closed","stream_id":"4ns1dtmm"}
-{"time":"2026-04-06T14:06:49.547065389+08:00","level":"INFO","msg":"stream: closed","id":"4ns1dtmm"}

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/debug.log

@@ -1,27 +0,0 @@
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Configure stats pid to 337747
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/train.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./train.py'}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Applying login settings: {}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug.log
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug-internal.log
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:init():619] calling init triggers
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
-config: {'name': 'train_bdpo_unet_lowdim', '_target_': 'llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace', 'obs_dim': 46, 'action_dim': 26, 'task_name': 'adroit-hand-hammer-v1', 'exp_name': 'default', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'policy': {'_target_': 'llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 26, 'local_cond_dim': None, 'global_cond_dim': 46, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'obs_dim': 46, 'action_dim': 26, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'hidden_size': [256, 256], 'activation': 'relu', 'gamma': 0.99, 'tau': 0.005, 'eta': 1.0, 'rho': 1.0, 'target_reduction': 'min', 'max_q_backup': False, 'actor_update_interval': 1, 'value_warmup_steps': 1000, 'num_eval_sample_steps': 100, 'act_limit': None, 'squash_action': False, 'init_weight': 0.01, 'checkpoint_path': None, 'behavior_checkpoint_path': 'data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0055-val_loss=0.215.ckpt', 'init_actor_from_behavior': True}, 'dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'critic_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'value_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'training': {'device': 'cuda:0', 'seed': 45, 'debug': False, 'resume': False, 'lr_scheduler': 'constant', 'lr_warmup_steps': 500, 'num_epochs': 201, 'gradient_accumulate_every': 1, 'grad_norm_clip': 5.0, 'rollout_every': 10, 'checkpoint_every': 10, 'val_every': 1, 'sample_every': 10, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'adroit-hand-hammer-v1-bdpo-training', 'resume': True, 'mode': 'online', 'name': '2026.04.05-19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'tags': ['train_bdpo_unet_lowdim', 'adroit-hand-hammer-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.04.05-19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}, 'task': {'name': 'adroit-hand-hammer-v1', 'obs_dim': 46, 'action_dim': 26, 'env_runner': {'_target_': 'llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner', 'env_name': 'llf-adroit-adroit-hand-hammer-v1', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 150, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset', 'data_path': 'datasets/adroit-hand-hammer-v1-general-reward-trans.pt', 'data_path2': 'datasets/adroit-hand-hammer-v1-reward-trans.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor'}}}
-2026-04-05 19:47:37,704 INFO    MainThread:337747 [wandb_init.py:init():669] starting backend
-2026-04-05 19:47:37,704 INFO    MainThread:337747 [wandb_init.py:init():673] sending inform_init request
-2026-04-05 19:47:37,706 INFO    MainThread:337747 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
-2026-04-05 19:47:37,707 INFO    MainThread:337747 [wandb_init.py:init():686] backend started and connected
-2026-04-05 19:47:37,722 INFO    MainThread:337747 [wandb_init.py:init():781] updated telemetry
-2026-04-05 19:47:37,839 INFO    MainThread:337747 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
-2026-04-05 19:47:38,798 INFO    MainThread:337747 [wandb_init.py:init():867] starting run threads in backend
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_console_start():2451] atexit reg
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_redirect():2299] redirect: wrap_raw
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_redirect():2364] Wrapping output streams.
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_redirect():2389] Redirects installed.
-2026-04-05 19:47:39,357 INFO    MainThread:337747 [wandb_init.py:init():911] run started, returning control to user process
-2026-04-05 19:47:39,358 INFO    MainThread:337747 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}
-2026-04-06 14:06:45,703 WARNING MsgRouterThr:337747 [router.py:message_loop():75] message_loop has been closed

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/files/config.yaml

@@ -1,250 +0,0 @@
-_target_:
-    value: llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace
-_wandb:
-    value:
-        cli_version: 0.18.6
-        m: []
-        python_version: 3.9.21
-        t:
-            "1":
-                - 1
-                - 2
-                - 3
-                - 5
-                - 11
-                - 12
-                - 41
-                - 49
-                - 50
-                - 51
-                - 53
-                - 55
-                - 71
-                - 83
-                - 95
-                - 98
-                - 100
-                - 105
-            "2":
-                - 1
-                - 2
-                - 3
-                - 5
-                - 11
-                - 12
-                - 41
-                - 49
-                - 50
-                - 51
-                - 53
-                - 55
-                - 71
-                - 83
-                - 95
-                - 98
-                - 100
-                - 105
-            "3":
-                - 13
-                - 15
-                - 16
-                - 23
-                - 55
-                - 61
-            "4": 3.9.21
-            "5": 0.18.6
-            "6": 4.47.1
-            "8":
-                - 5
-            "12": 0.18.6
-            "13": linux-x86_64
-action_dim:
-    value: 26
-checkpoint:
-    value:
-        save_last_ckpt: true
-        save_last_snapshot: false
-        topk:
-            format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
-            k: 5
-            mode: max
-            monitor_key: test_success_rate
-critic_optimizer:
-    value:
-        _target_: torch.optim.Adam
-        betas:
-            - 0.9
-            - 0.999
-        eps: 1e-08
-        lr: 0.0001
-        weight_decay: 0
-dataloader:
-    value:
-        batch_size: 256
-        num_workers: 0
-        persistent_workers: false
-        pin_memory: false
-        shuffle: true
-exp_name:
-    value: default
-horizon:
-    value: 1
-logging:
-    value:
-        group: null
-        id: null
-        mode: online
-        name: 2026.04.05-19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-        project: adroit-hand-hammer-v1-bdpo-training
-        resume: true
-        tags:
-            - train_bdpo_unet_lowdim
-            - adroit-hand-hammer-v1
-            - default
-multi_run:
-    value:
-        run_dir: data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-        wandb_name_base: 2026.04.05-19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-n_action_steps:
-    value: 1
-n_latency_steps:
-    value: 0
-n_obs_steps:
-    value: 1
-name:
-    value: train_bdpo_unet_lowdim
-obs_dim:
-    value: 46
-optimizer:
-    value:
-        _target_: torch.optim.Adam
-        betas:
-            - 0.9
-            - 0.999
-        eps: 1e-08
-        lr: 0.0001
-        weight_decay: 0
-output_dir:
-    value: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-past_action_visible:
-    value: false
-policy:
-    value:
-        _target_: llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy
-        act_limit: null
-        action_dim: 26
-        activation: relu
-        actor_update_interval: 1
-        behavior_checkpoint_path: data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0055-val_loss=0.215.ckpt
-        checkpoint_path: null
-        eta: 1
-        gamma: 0.99
-        hidden_size:
-            - 256
-            - 256
-        horizon: 1
-        init_actor_from_behavior: true
-        init_weight: 0.01
-        max_q_backup: false
-        model:
-            _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
-            cond_predict_scale: true
-            diffusion_step_embed_dim: 256
-            down_dims:
-                - 256
-                - 512
-                - 1024
-            global_cond_dim: 46
-            input_dim: 26
-            kernel_size: 5
-            local_cond_dim: null
-            n_groups: 8
-        n_action_steps: 1
-        n_obs_steps: 1
-        noise_scheduler:
-            _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
-            beta_end: 0.02
-            beta_schedule: squaredcos_cap_v2
-            beta_start: 0.0001
-            clip_sample: true
-            num_train_timesteps: 100
-            prediction_type: epsilon
-            variance_type: fixed_small
-        num_eval_sample_steps: 100
-        obs_dim: 46
-        rho: 1
-        squash_action: false
-        target_reduction: min
-        tau: 0.005
-        value_warmup_steps: 1000
-task:
-    value:
-        action_dim: 26
-        dataset:
-            _target_: llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset
-            data_path: datasets/adroit-hand-hammer-v1-general-reward-trans.pt
-            data_path2: datasets/adroit-hand-hammer-v1-reward-trans.pt
-            dummy_normalizer: false
-            horizon: 1
-            obs_eef_target: true
-            pad_after: 0
-            pad_before: 0
-            use_manual_normalizer: false
-            val_ratio: 0.1
-        env_runner:
-            _target_: llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner
-            discount: 0.99
-            env_name: llf-adroit-adroit-hand-hammer-v1
-            feedback_type:
-                - hp
-                - hn
-                - fp
-            instruction_type: b
-            max_steps: 150
-            n_action_steps: 1
-            n_envs: 10
-            n_obs_steps: 1
-            n_test: 50
-            n_train: 10
-            visual: false
-        instructor:
-            _target_: llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor
-        name: adroit-hand-hammer-v1
-        obs_dim: 46
-task_name:
-    value: adroit-hand-hammer-v1
-training:
-    value:
-        checkpoint_every: 10
-        debug: false
-        device: cuda:0
-        grad_norm_clip: 5
-        gradient_accumulate_every: 1
-        lr_scheduler: constant
-        lr_warmup_steps: 500
-        max_train_steps: null
-        max_val_steps: null
-        num_epochs: 201
-        resume: false
-        rollout_every: 10
-        sample_every: 10
-        sample_max_batch: 128
-        seed: 45
-        tqdm_interval_sec: 1
-        val_every: 1
-val_dataloader:
-    value:
-        batch_size: 256
-        num_workers: 0
-        persistent_workers: false
-        pin_memory: false
-        shuffle: true
-value_optimizer:
-    value:
-        _target_: torch.optim.Adam
-        betas:
-            - 0.9
-            - 0.999
-        eps: 1e-08
-        lr: 0.0001
-        weight_decay: 0

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/files/output.log

@@ -1,3 +0,0 @@
-Eval AdroitHandLowdimRunner 1/6:   0%|                                                                                                             | 0/150 [00:00<?, ?it/s]/tmp2/chyang/workspace/LLM-BC/llmbc/common/llfbench_util.py:39: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:275.)
-  obs = torch.tensor(obs, dtype=torch.float32).unsqueeze(dim=0).to(device)
-                                                                                                                                                                           

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/files/wandb-metadata.json

@@ -1,57 +0,0 @@
-{
-  "os": "Linux-6.8.0-90-generic-x86_64-with-glibc2.35",
-  "python": "3.9.21",
-  "startedAt": "2026-04-05T11:47:37.707689Z",
-  "args": [
-    "--config-path",
-    "config/main_table",
-    "--config-name",
-    "bdpo_adroit-hand-hammer-v1_gen.yaml",
-    "training.seed=45",
-    "policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\\=0055-val_loss\\=0.215.ckpt"
-  ],
-  "program": "/tmp2/chyang/workspace/LLM-BC/./train.py",
-  "codePath": "train.py",
-  "git": {
-    "remote": "https://github.com/CHYang25/LLM-BC.git",
-    "commit": "a983d49eed9d3b7c64b22a85eb6908c8694405af"
-  },
-  "email": "chris920325@gmail.com",
-  "root": "/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1",
-  "host": "rllab1012",
-  "username": "chyang",
-  "executable": "/home/chyang/miniconda3/envs/llm-bc/bin/python3",
-  "codePathLocal": "train.py",
-  "cpu_count": 10,
-  "cpu_count_logical": 20,
-  "gpu": "NVIDIA GeForce RTX 4070 Ti",
-  "gpu_count": 2,
-  "disk": {
-    "/": {
-      "total": "982820896768",
-      "used": "414740176896"
-    }
-  },
-  "memory": {
-    "total": "134736023552"
-  },
-  "cpu": {
-    "count": 10,
-    "countLogical": 20
-  },
-  "gpu_nvidia": [
-    {
-      "name": "NVIDIA GeForce RTX 4070 Ti",
-      "memoryTotal": "12878610432",
-      "cudaCores": 7680,
-      "architecture": "Ada"
-    },
-    {
-      "name": "NVIDIA GeForce RTX 4070 Ti",
-      "memoryTotal": "12878610432",
-      "cudaCores": 7680,
-      "architecture": "Ada"
-    }
-  ],
-  "cudaVersion": "13.0"
-}

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/files/wandb-summary.json

@@ -1 +0,0 @@
-{"stats/bootstrap_mean":-132.64468383789062,"stats/t_mean":48.97916793823242,"test/mean_score":-1.2400550559828623,"_wandb":{"runtime":65947},"stats/done_mean":0.010416666977107525,"loss/v1":11.482786178588867,"loss/q2":371.92071533203125,"stats/reward_mean":0.5011574029922485,"train/success_rate":0,"global_step":17687,"loss/v":26.77777862548828,"test/success_rate":0,"train_loss":5260.534036042,"_runtime":65947.995279549,"epoch":200,"loss/q":769.3199462890625,"lr":0.0001,"val_loss":5150.252891975641,"train/mean_score":-1.2464290659121704,"loss/actor_bootstrap":-132.64468383789062,"loss/q1":397.3992614746094,"_timestamp":1.775455588772426e+09,"loss/actor":133.28306579589844,"_step":17687,"train/cumulative_reward":-103.11810478313211,"loss/v2":15.29499340057373,"test/cumulative_reward":-102.2961005109052,"loss/actor_kl":0.6383782029151917,"stats/kl_mean":0.6383782029151917}

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug-core.log

@@ -1,14 +0,0 @@
-{"time":"2026-04-05T19:47:37.00983905+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpol07t4o4/port-337747.txt","pid":337747,"debug":false,"disable-analytics":false}
-{"time":"2026-04-05T19:47:37.009863037+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
-{"time":"2026-04-05T19:47:37.027426037+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":337747}
-{"time":"2026-04-05T19:47:37.027459487+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":45777,"Zone":""}}
-{"time":"2026-04-05T19:47:37.206359992+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:51182"}
-{"time":"2026-04-05T19:47:37.71135652+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"4ns1dtmm","id":"127.0.0.1:51182"}
-{"time":"2026-04-05T19:47:37.820725434+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"4ns1dtmm","id":"127.0.0.1:51182"}
-{"time":"2026-04-06T14:06:45.702830394+08:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:51182"}
-{"time":"2026-04-06T14:06:45.703001947+08:00","level":"INFO","msg":"server is shutting down"}
-{"time":"2026-04-06T14:06:45.702969429+08:00","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:51182"}
-{"time":"2026-04-06T14:06:45.70315781+08:00","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:51182"}
-{"time":"2026-04-06T14:06:49.547222645+08:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:51182"}
-{"time":"2026-04-06T14:06:49.547299852+08:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:51182"}
-{"time":"2026-04-06T14:06:49.547332331+08:00","level":"INFO","msg":"server is closed"}

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug-internal.log

@@ -1,16 +0,0 @@
-{"time":"2026-04-05T19:47:37.711575744+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
-{"time":"2026-04-05T19:47:37.711598693+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug-core.log"}
-{"time":"2026-04-05T19:47:37.820659395+08:00","level":"INFO","msg":"created new stream","id":"4ns1dtmm"}
-{"time":"2026-04-05T19:47:37.820714248+08:00","level":"INFO","msg":"stream: started","id":"4ns1dtmm"}
-{"time":"2026-04-05T19:47:37.820739406+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-05T19:47:37.820815969+08:00","level":"INFO","msg":"sender: started","stream_id":"4ns1dtmm"}
-{"time":"2026-04-05T19:47:37.820853585+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-05T19:47:38.804274688+08:00","level":"INFO","msg":"Starting system monitor"}
-{"time":"2026-04-06T14:06:45.702933515+08:00","level":"INFO","msg":"stream: closing","id":"4ns1dtmm"}
-{"time":"2026-04-06T14:06:45.702982796+08:00","level":"INFO","msg":"Stopping system monitor"}
-{"time":"2026-04-06T14:06:45.703854988+08:00","level":"INFO","msg":"Stopped system monitor"}
-{"time":"2026-04-06T14:06:49.250748912+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
-{"time":"2026-04-06T14:06:49.546933574+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-06T14:06:49.546991191+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"4ns1dtmm"}}
-{"time":"2026-04-06T14:06:49.547007286+08:00","level":"INFO","msg":"sender: closed","stream_id":"4ns1dtmm"}
-{"time":"2026-04-06T14:06:49.547065389+08:00","level":"INFO","msg":"stream: closed","id":"4ns1dtmm"}

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug.log

@@ -1,27 +0,0 @@
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Configure stats pid to 337747
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
-2026-04-05 19:47:37,702 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/train.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./train.py'}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_setup.py:_flush():79] Applying login settings: {}
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug.log
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/logs/debug-internal.log
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:init():619] calling init triggers
-2026-04-05 19:47:37,703 INFO    MainThread:337747 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
-config: {'name': 'train_bdpo_unet_lowdim', '_target_': 'llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace', 'obs_dim': 46, 'action_dim': 26, 'task_name': 'adroit-hand-hammer-v1', 'exp_name': 'default', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'policy': {'_target_': 'llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 26, 'local_cond_dim': None, 'global_cond_dim': 46, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'obs_dim': 46, 'action_dim': 26, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'hidden_size': [256, 256], 'activation': 'relu', 'gamma': 0.99, 'tau': 0.005, 'eta': 1.0, 'rho': 1.0, 'target_reduction': 'min', 'max_q_backup': False, 'actor_update_interval': 1, 'value_warmup_steps': 1000, 'num_eval_sample_steps': 100, 'act_limit': None, 'squash_action': False, 'init_weight': 0.01, 'checkpoint_path': None, 'behavior_checkpoint_path': 'data/outputs/2026.04.04/12.00.56_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0055-val_loss=0.215.ckpt', 'init_actor_from_behavior': True}, 'dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'critic_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'value_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'training': {'device': 'cuda:0', 'seed': 45, 'debug': False, 'resume': False, 'lr_scheduler': 'constant', 'lr_warmup_steps': 500, 'num_epochs': 201, 'gradient_accumulate_every': 1, 'grad_norm_clip': 5.0, 'rollout_every': 10, 'checkpoint_every': 10, 'val_every': 1, 'sample_every': 10, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'adroit-hand-hammer-v1-bdpo-training', 'resume': True, 'mode': 'online', 'name': '2026.04.05-19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'tags': ['train_bdpo_unet_lowdim', 'adroit-hand-hammer-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.04.05-19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}, 'task': {'name': 'adroit-hand-hammer-v1', 'obs_dim': 46, 'action_dim': 26, 'env_runner': {'_target_': 'llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner', 'env_name': 'llf-adroit-adroit-hand-hammer-v1', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 150, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset', 'data_path': 'datasets/adroit-hand-hammer-v1-general-reward-trans.pt', 'data_path2': 'datasets/adroit-hand-hammer-v1-reward-trans.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor'}}}
-2026-04-05 19:47:37,704 INFO    MainThread:337747 [wandb_init.py:init():669] starting backend
-2026-04-05 19:47:37,704 INFO    MainThread:337747 [wandb_init.py:init():673] sending inform_init request
-2026-04-05 19:47:37,706 INFO    MainThread:337747 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
-2026-04-05 19:47:37,707 INFO    MainThread:337747 [wandb_init.py:init():686] backend started and connected
-2026-04-05 19:47:37,722 INFO    MainThread:337747 [wandb_init.py:init():781] updated telemetry
-2026-04-05 19:47:37,839 INFO    MainThread:337747 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
-2026-04-05 19:47:38,798 INFO    MainThread:337747 [wandb_init.py:init():867] starting run threads in backend
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_console_start():2451] atexit reg
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_redirect():2299] redirect: wrap_raw
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_redirect():2364] Wrapping output streams.
-2026-04-05 19:47:39,355 INFO    MainThread:337747 [wandb_run.py:_redirect():2389] Redirects installed.
-2026-04-05 19:47:39,357 INFO    MainThread:337747 [wandb_init.py:init():911] run started, returning control to user process
-2026-04-05 19:47:39,358 INFO    MainThread:337747 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}
-2026-04-06 14:06:45,703 WARNING MsgRouterThr:337747 [router.py:message_loop():75] message_loop has been closed

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_194737-4ns1dtmm/run-4ns1dtmm.wandb

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:aa79c191fcbc824eb60ac297cc1c938864b9b4b3cc644464f6875d37f27750c1
-size 51380340

2026.04.05/19.47.17_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/wandb-resume.json

@@ -1 +0,0 @@
-{"run_id": "4ns1dtmm"}

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/.hydra/config.yaml

@@ -1,168 +0,0 @@
-name: train_bdpo_unet_lowdim
-_target_: llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace
-obs_dim: ${task.obs_dim}
-action_dim: ${task.action_dim}
-task_name: ${task.name}
-exp_name: default
-horizon: 1
-n_obs_steps: 1
-n_action_steps: 1
-n_latency_steps: 0
-past_action_visible: false
-policy:
-  _target_: llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy
-  model:
-    _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
-    input_dim: ${task.action_dim}
-    local_cond_dim: null
-    global_cond_dim: '${eval: ${task.obs_dim}*${n_obs_steps}}'
-    diffusion_step_embed_dim: 256
-    down_dims:
-    - 256
-    - 512
-    - 1024
-    kernel_size: 5
-    n_groups: 8
-    cond_predict_scale: true
-  noise_scheduler:
-    _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
-    num_train_timesteps: 100
-    beta_start: 0.0001
-    beta_end: 0.02
-    beta_schedule: squaredcos_cap_v2
-    variance_type: fixed_small
-    clip_sample: true
-    prediction_type: epsilon
-  obs_dim: ${obs_dim}
-  action_dim: ${action_dim}
-  horizon: ${horizon}
-  n_obs_steps: ${n_obs_steps}
-  n_action_steps: ${n_action_steps}
-  hidden_size:
-  - 256
-  - 256
-  activation: relu
-  gamma: ${task.env_runner.discount}
-  tau: 0.005
-  eta: 1.0
-  rho: 1.0
-  target_reduction: min
-  max_q_backup: false
-  actor_update_interval: 1
-  value_warmup_steps: 1000
-  num_eval_sample_steps: 100
-  act_limit: null
-  squash_action: false
-  init_weight: 0.01
-  checkpoint_path: null
-  behavior_checkpoint_path: data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0095-val_loss=0.199.ckpt
-  init_actor_from_behavior: true
-dataloader:
-  batch_size: 256
-  num_workers: 0
-  shuffle: true
-  pin_memory: false
-  persistent_workers: false
-val_dataloader:
-  batch_size: 256
-  num_workers: 0
-  shuffle: true
-  pin_memory: false
-  persistent_workers: false
-optimizer:
-  _target_: torch.optim.Adam
-  lr: 0.0001
-  betas:
-  - 0.9
-  - 0.999
-  eps: 1.0e-08
-  weight_decay: 0.0
-critic_optimizer:
-  _target_: torch.optim.Adam
-  lr: 0.0001
-  betas:
-  - 0.9
-  - 0.999
-  eps: 1.0e-08
-  weight_decay: 0.0
-value_optimizer:
-  _target_: torch.optim.Adam
-  lr: 0.0001
-  betas:
-  - 0.9
-  - 0.999
-  eps: 1.0e-08
-  weight_decay: 0.0
-training:
-  device: cuda:0
-  seed: 46
-  debug: false
-  resume: false
-  lr_scheduler: constant
-  lr_warmup_steps: 500
-  num_epochs: 201
-  gradient_accumulate_every: 1
-  grad_norm_clip: 5.0
-  rollout_every: 10
-  checkpoint_every: 10
-  val_every: 1
-  sample_every: 10
-  sample_max_batch: 128
-  max_train_steps: null
-  max_val_steps: null
-  tqdm_interval_sec: 1.0
-logging:
-  project: ${task.name}-bdpo-training
-  resume: true
-  mode: online
-  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
-  tags:
-  - ${name}
-  - ${task_name}
-  - ${exp_name}
-  id: null
-  group: null
-checkpoint:
-  topk:
-    monitor_key: test_success_rate
-    mode: max
-    k: 5
-    format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
-  save_last_ckpt: true
-  save_last_snapshot: false
-multi_run:
-  run_dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
-  wandb_name_base: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
-task:
-  name: adroit-hand-hammer-v1
-  obs_dim: 46
-  action_dim: 26
-  env_runner:
-    _target_: llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner
-    env_name: llf-adroit-adroit-hand-hammer-v1
-    n_train: 10
-    n_test: 50
-    n_envs: 10
-    max_steps: 150
-    n_obs_steps: ${n_obs_steps}
-    n_action_steps: ${n_action_steps}
-    instruction_type: b
-    feedback_type:
-    - hp
-    - hn
-    - fp
-    visual: false
-    discount: 0.99
-  dataset:
-    _target_: llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset
-    data_path: datasets/adroit-hand-hammer-v1-general-reward-trans.pt
-    data_path2: datasets/adroit-hand-hammer-v1-reward-trans.pt
-    horizon: ${horizon}
-    pad_before: ${eval:'${n_obs_steps}-1'}
-    pad_after: ${eval:'${n_action_steps}-1'}
-    obs_eef_target: true
-    use_manual_normalizer: false
-    val_ratio: 0.1
-    dummy_normalizer: false
-  instructor:
-    _target_: llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/.hydra/hydra.yaml

@@ -1,156 +0,0 @@
-hydra:
-  run:
-    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
-  sweep:
-    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
-    subdir: ${hydra.job.num}
-  launcher:
-    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
-  sweeper:
-    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
-    max_batch_size: null
-    params: null
-  help:
-    app_name: ${hydra.job.name}
-    header: '${hydra.help.app_name} is powered by Hydra.
-
-      '
-    footer: 'Powered by Hydra (https://hydra.cc)
-
-      Use --hydra-help to view Hydra specific help
-
-      '
-    template: '${hydra.help.header}
-
-      == Configuration groups ==
-
-      Compose your configuration from those groups (group=option)
-
-
-      $APP_CONFIG_GROUPS
-
-
-      == Config ==
-
-      Override anything in the config (foo.bar=value)
-
-
-      $CONFIG
-
-
-      ${hydra.help.footer}
-
-      '
-  hydra_help:
-    template: 'Hydra (${hydra.runtime.version})
-
-      See https://hydra.cc for more info.
-
-
-      == Flags ==
-
-      $FLAGS_HELP
-
-
-      == Configuration groups ==
-
-      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
-      to command line)
-
-
-      $HYDRA_CONFIG_GROUPS
-
-
-      Use ''--cfg hydra'' to Show the Hydra config.
-
-      '
-    hydra_help: ???
-  hydra_logging:
-    version: 1
-    formatters:
-      simple:
-        format: '[%(asctime)s][HYDRA] %(message)s'
-    handlers:
-      console:
-        class: logging.StreamHandler
-        formatter: simple
-        stream: ext://sys.stdout
-    root:
-      level: INFO
-      handlers:
-      - console
-    loggers:
-      logging_example:
-        level: DEBUG
-    disable_existing_loggers: false
-  job_logging:
-    version: 1
-    formatters:
-      simple:
-        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
-    handlers:
-      console:
-        class: logging.StreamHandler
-        formatter: simple
-        stream: ext://sys.stdout
-      file:
-        class: logging.FileHandler
-        formatter: simple
-        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
-    root:
-      level: INFO
-      handlers:
-      - console
-      - file
-    disable_existing_loggers: false
-  env: {}
-  mode: RUN
-  searchpath: []
-  callbacks: {}
-  output_subdir: .hydra
-  overrides:
-    hydra:
-    - hydra.mode=RUN
-    task:
-    - training.seed=46
-    - policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\=0095-val_loss\=0.199.ckpt
-  job:
-    name: train
-    chdir: null
-    override_dirname: policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\=0095-val_loss\=0.199.ckpt,training.seed=46
-    id: ???
-    num: ???
-    config_name: bdpo_adroit-hand-hammer-v1_gen.yaml
-    env_set: {}
-    env_copy: []
-    config:
-      override_dirname:
-        kv_sep: '='
-        item_sep: ','
-        exclude_keys: []
-  runtime:
-    version: 1.2.0
-    version_base: '1.2'
-    cwd: /tmp2/chyang/workspace/LLM-BC
-    config_sources:
-    - path: hydra.conf
-      schema: pkg
-      provider: hydra
-    - path: /tmp2/chyang/workspace/LLM-BC/config/main_table
-      schema: file
-      provider: main
-    - path: ''
-      schema: structured
-      provider: schema
-    output_dir: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-    choices:
-      hydra/env: default
-      hydra/callbacks: null
-      hydra/job_logging: default
-      hydra/hydra_logging: default
-      hydra/hydra_help: default
-      hydra/help: default
-      hydra/sweeper: basic
-      hydra/launcher: basic
-      hydra/output: default
-  verbose: false

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/.hydra/overrides.yaml

@@ -1,2 +0,0 @@
-- training.seed=46
-- policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\=0095-val_loss\=0.199.ckpt

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0000-test_success_rate=0.260.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:47079e0e4d5b09e26e06694447130ac329307e1f6b535571b9b94c1d2280f0f7
-size 1055778720

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0010-test_success_rate=0.380.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:40bbe1b7b98ea2387dba707176f101aebba65d970e3e460dc02953a66d671fc8
-size 1055778720

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0020-test_success_rate=0.160.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2480d5100d2f36645ffe097e9c3cb3d8c8dab192e369784b5c61b3cad8a6d23b
-size 1580582693

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0030-test_success_rate=0.360.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:919839d7af703a49b8938af2dda0f44953ac2eb5ed3829aaa72751afec860800
-size 1580582693

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0040-test_success_rate=0.300.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:9b4bf2991dfc5b6c2eefa93a5ba4a9b1c5bfde1595f9542aa60e767780e758b2
-size 1580582693

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/checkpoints/latest.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:aeaadcac1a542e9eb381a8982b030d9d0f4a0b4bb752fa4dac6d97b92a05c79e
-size 1580582693

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/train.log

@@ -1,11 +0,0 @@
-[2026-04-05 20:14:52,028][numexpr.utils][INFO] - Note: NumExpr detected 20 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
-[2026-04-05 20:14:52,028][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
-[2026-04-05 20:14:52,509][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.558467e+07
-[2026-04-05 20:14:56,299][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.558467e+07
-[2026-04-05 20:14:59,375][OpenGL.acceleratesupport][WARNING] - Incompatible version of OpenGL_accelerate found, need at least (3, 1, 10) found (3, 1, 9)
-[2026-04-05 20:14:59,375][OpenGL.acceleratesupport][INFO] - No OpenGL_accelerate module loaded: Old version of OpenGL_accelerate
-[2026-04-05 20:15:01,473][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
-[2026-04-05 20:15:01,479][absl][INFO] - MuJoCo library version is: 2.3.7
-[2026-04-05 20:15:01,580][datasets][INFO] - PyTorch version 2.2.2 available.
-[2026-04-05 20:15:01,581][datasets][INFO] - TensorFlow version 2.15.1 available.
-[2026-04-05 20:15:01,581][datasets][INFO] - JAX version 0.4.30 available.

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/debug-internal.log

@@ -1,16 +0,0 @@
-{"time":"2026-04-05T20:15:02.675207147+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
-{"time":"2026-04-05T20:15:02.675236072+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug-core.log"}
-{"time":"2026-04-05T20:15:02.783639967+08:00","level":"INFO","msg":"created new stream","id":"gsxv5o2x"}
-{"time":"2026-04-05T20:15:02.783682009+08:00","level":"INFO","msg":"stream: started","id":"gsxv5o2x"}
-{"time":"2026-04-05T20:15:02.783719383+08:00","level":"INFO","msg":"sender: started","stream_id":"gsxv5o2x"}
-{"time":"2026-04-05T20:15:02.783704075+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-05T20:15:02.783701477+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-05T20:15:03.681306658+08:00","level":"INFO","msg":"Starting system monitor"}
-{"time":"2026-04-06T14:21:13.810659156+08:00","level":"INFO","msg":"stream: closing","id":"gsxv5o2x"}
-{"time":"2026-04-06T14:21:13.81070755+08:00","level":"INFO","msg":"Stopping system monitor"}
-{"time":"2026-04-06T14:21:13.811296491+08:00","level":"INFO","msg":"Stopped system monitor"}
-{"time":"2026-04-06T14:21:15.381579169+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
-{"time":"2026-04-06T14:21:15.820068161+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-06T14:21:15.820142874+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-06T14:21:15.820175329+08:00","level":"INFO","msg":"sender: closed","stream_id":"gsxv5o2x"}
-{"time":"2026-04-06T14:21:15.820335911+08:00","level":"INFO","msg":"stream: closed","id":"gsxv5o2x"}

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/debug.log

@@ -1,27 +0,0 @@
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Configure stats pid to 349699
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/train.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./train.py'}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Applying login settings: {}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug.log
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug-internal.log
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:init():619] calling init triggers
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
-config: {'name': 'train_bdpo_unet_lowdim', '_target_': 'llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace', 'obs_dim': 46, 'action_dim': 26, 'task_name': 'adroit-hand-hammer-v1', 'exp_name': 'default', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'policy': {'_target_': 'llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 26, 'local_cond_dim': None, 'global_cond_dim': 46, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'obs_dim': 46, 'action_dim': 26, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'hidden_size': [256, 256], 'activation': 'relu', 'gamma': 0.99, 'tau': 0.005, 'eta': 1.0, 'rho': 1.0, 'target_reduction': 'min', 'max_q_backup': False, 'actor_update_interval': 1, 'value_warmup_steps': 1000, 'num_eval_sample_steps': 100, 'act_limit': None, 'squash_action': False, 'init_weight': 0.01, 'checkpoint_path': None, 'behavior_checkpoint_path': 'data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0095-val_loss=0.199.ckpt', 'init_actor_from_behavior': True}, 'dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'critic_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'value_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'training': {'device': 'cuda:0', 'seed': 46, 'debug': False, 'resume': False, 'lr_scheduler': 'constant', 'lr_warmup_steps': 500, 'num_epochs': 201, 'gradient_accumulate_every': 1, 'grad_norm_clip': 5.0, 'rollout_every': 10, 'checkpoint_every': 10, 'val_every': 1, 'sample_every': 10, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'adroit-hand-hammer-v1-bdpo-training', 'resume': True, 'mode': 'online', 'name': '2026.04.05-20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'tags': ['train_bdpo_unet_lowdim', 'adroit-hand-hammer-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.04.05-20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}, 'task': {'name': 'adroit-hand-hammer-v1', 'obs_dim': 46, 'action_dim': 26, 'env_runner': {'_target_': 'llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner', 'env_name': 'llf-adroit-adroit-hand-hammer-v1', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 150, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset', 'data_path': 'datasets/adroit-hand-hammer-v1-general-reward-trans.pt', 'data_path2': 'datasets/adroit-hand-hammer-v1-reward-trans.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor'}}}
-2026-04-05 20:15:02,669 INFO    MainThread:349699 [wandb_init.py:init():669] starting backend
-2026-04-05 20:15:02,669 INFO    MainThread:349699 [wandb_init.py:init():673] sending inform_init request
-2026-04-05 20:15:02,671 INFO    MainThread:349699 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
-2026-04-05 20:15:02,671 INFO    MainThread:349699 [wandb_init.py:init():686] backend started and connected
-2026-04-05 20:15:02,678 INFO    MainThread:349699 [wandb_init.py:init():781] updated telemetry
-2026-04-05 20:15:02,741 INFO    MainThread:349699 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
-2026-04-05 20:15:03,675 INFO    MainThread:349699 [wandb_init.py:init():867] starting run threads in backend
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_console_start():2451] atexit reg
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_redirect():2299] redirect: wrap_raw
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_redirect():2364] Wrapping output streams.
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_redirect():2389] Redirects installed.
-2026-04-05 20:15:04,116 INFO    MainThread:349699 [wandb_init.py:init():911] run started, returning control to user process
-2026-04-05 20:15:04,117 INFO    MainThread:349699 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}
-2026-04-06 14:21:13,810 WARNING MsgRouterThr:349699 [router.py:message_loop():75] message_loop has been closed

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/files/config.yaml

@@ -1,250 +0,0 @@
-_target_:
-    value: llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace
-_wandb:
-    value:
-        cli_version: 0.18.6
-        m: []
-        python_version: 3.9.21
-        t:
-            "1":
-                - 1
-                - 2
-                - 3
-                - 5
-                - 11
-                - 12
-                - 41
-                - 49
-                - 50
-                - 51
-                - 53
-                - 55
-                - 71
-                - 83
-                - 95
-                - 98
-                - 100
-                - 105
-            "2":
-                - 1
-                - 2
-                - 3
-                - 5
-                - 11
-                - 12
-                - 41
-                - 49
-                - 50
-                - 51
-                - 53
-                - 55
-                - 71
-                - 83
-                - 95
-                - 98
-                - 100
-                - 105
-            "3":
-                - 13
-                - 15
-                - 16
-                - 23
-                - 55
-                - 61
-            "4": 3.9.21
-            "5": 0.18.6
-            "6": 4.47.1
-            "8":
-                - 5
-            "12": 0.18.6
-            "13": linux-x86_64
-action_dim:
-    value: 26
-checkpoint:
-    value:
-        save_last_ckpt: true
-        save_last_snapshot: false
-        topk:
-            format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
-            k: 5
-            mode: max
-            monitor_key: test_success_rate
-critic_optimizer:
-    value:
-        _target_: torch.optim.Adam
-        betas:
-            - 0.9
-            - 0.999
-        eps: 1e-08
-        lr: 0.0001
-        weight_decay: 0
-dataloader:
-    value:
-        batch_size: 256
-        num_workers: 0
-        persistent_workers: false
-        pin_memory: false
-        shuffle: true
-exp_name:
-    value: default
-horizon:
-    value: 1
-logging:
-    value:
-        group: null
-        id: null
-        mode: online
-        name: 2026.04.05-20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-        project: adroit-hand-hammer-v1-bdpo-training
-        resume: true
-        tags:
-            - train_bdpo_unet_lowdim
-            - adroit-hand-hammer-v1
-            - default
-multi_run:
-    value:
-        run_dir: data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-        wandb_name_base: 2026.04.05-20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-n_action_steps:
-    value: 1
-n_latency_steps:
-    value: 0
-n_obs_steps:
-    value: 1
-name:
-    value: train_bdpo_unet_lowdim
-obs_dim:
-    value: 46
-optimizer:
-    value:
-        _target_: torch.optim.Adam
-        betas:
-            - 0.9
-            - 0.999
-        eps: 1e-08
-        lr: 0.0001
-        weight_decay: 0
-output_dir:
-    value: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1
-past_action_visible:
-    value: false
-policy:
-    value:
-        _target_: llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy
-        act_limit: null
-        action_dim: 26
-        activation: relu
-        actor_update_interval: 1
-        behavior_checkpoint_path: data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0095-val_loss=0.199.ckpt
-        checkpoint_path: null
-        eta: 1
-        gamma: 0.99
-        hidden_size:
-            - 256
-            - 256
-        horizon: 1
-        init_actor_from_behavior: true
-        init_weight: 0.01
-        max_q_backup: false
-        model:
-            _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
-            cond_predict_scale: true
-            diffusion_step_embed_dim: 256
-            down_dims:
-                - 256
-                - 512
-                - 1024
-            global_cond_dim: 46
-            input_dim: 26
-            kernel_size: 5
-            local_cond_dim: null
-            n_groups: 8
-        n_action_steps: 1
-        n_obs_steps: 1
-        noise_scheduler:
-            _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
-            beta_end: 0.02
-            beta_schedule: squaredcos_cap_v2
-            beta_start: 0.0001
-            clip_sample: true
-            num_train_timesteps: 100
-            prediction_type: epsilon
-            variance_type: fixed_small
-        num_eval_sample_steps: 100
-        obs_dim: 46
-        rho: 1
-        squash_action: false
-        target_reduction: min
-        tau: 0.005
-        value_warmup_steps: 1000
-task:
-    value:
-        action_dim: 26
-        dataset:
-            _target_: llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset
-            data_path: datasets/adroit-hand-hammer-v1-general-reward-trans.pt
-            data_path2: datasets/adroit-hand-hammer-v1-reward-trans.pt
-            dummy_normalizer: false
-            horizon: 1
-            obs_eef_target: true
-            pad_after: 0
-            pad_before: 0
-            use_manual_normalizer: false
-            val_ratio: 0.1
-        env_runner:
-            _target_: llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner
-            discount: 0.99
-            env_name: llf-adroit-adroit-hand-hammer-v1
-            feedback_type:
-                - hp
-                - hn
-                - fp
-            instruction_type: b
-            max_steps: 150
-            n_action_steps: 1
-            n_envs: 10
-            n_obs_steps: 1
-            n_test: 50
-            n_train: 10
-            visual: false
-        instructor:
-            _target_: llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor
-        name: adroit-hand-hammer-v1
-        obs_dim: 46
-task_name:
-    value: adroit-hand-hammer-v1
-training:
-    value:
-        checkpoint_every: 10
-        debug: false
-        device: cuda:0
-        grad_norm_clip: 5
-        gradient_accumulate_every: 1
-        lr_scheduler: constant
-        lr_warmup_steps: 500
-        max_train_steps: null
-        max_val_steps: null
-        num_epochs: 201
-        resume: false
-        rollout_every: 10
-        sample_every: 10
-        sample_max_batch: 128
-        seed: 46
-        tqdm_interval_sec: 1
-        val_every: 1
-val_dataloader:
-    value:
-        batch_size: 256
-        num_workers: 0
-        persistent_workers: false
-        pin_memory: false
-        shuffle: true
-value_optimizer:
-    value:
-        _target_: torch.optim.Adam
-        betas:
-            - 0.9
-            - 0.999
-        eps: 1e-08
-        lr: 0.0001
-        weight_decay: 0

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/files/output.log

@@ -1,3 +0,0 @@
-Eval AdroitHandLowdimRunner 1/6:   0%|                                                                                                             | 0/150 [00:00<?, ?it/s]/tmp2/chyang/workspace/LLM-BC/llmbc/common/llfbench_util.py:39: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:275.)
-  obs = torch.tensor(obs, dtype=torch.float32).unsqueeze(dim=0).to(device)
-                                                                                                                                                                           

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/files/wandb-metadata.json

@@ -1,57 +0,0 @@
-{
-  "os": "Linux-6.8.0-90-generic-x86_64-with-glibc2.35",
-  "python": "3.9.21",
-  "startedAt": "2026-04-05T12:15:02.672124Z",
-  "args": [
-    "--config-path",
-    "config/main_table",
-    "--config-name",
-    "bdpo_adroit-hand-hammer-v1_gen.yaml",
-    "training.seed=46",
-    "policy.behavior_checkpoint_path=data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch\\=0095-val_loss\\=0.199.ckpt"
-  ],
-  "program": "/tmp2/chyang/workspace/LLM-BC/./train.py",
-  "codePath": "train.py",
-  "git": {
-    "remote": "https://github.com/CHYang25/LLM-BC.git",
-    "commit": "a983d49eed9d3b7c64b22a85eb6908c8694405af"
-  },
-  "email": "chris920325@gmail.com",
-  "root": "/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1",
-  "host": "rllab1012",
-  "username": "chyang",
-  "executable": "/home/chyang/miniconda3/envs/llm-bc/bin/python3",
-  "codePathLocal": "train.py",
-  "cpu_count": 10,
-  "cpu_count_logical": 20,
-  "gpu": "NVIDIA GeForce RTX 4070 Ti",
-  "gpu_count": 2,
-  "disk": {
-    "/": {
-      "total": "982820896768",
-      "used": "414742159360"
-    }
-  },
-  "memory": {
-    "total": "134736023552"
-  },
-  "cpu": {
-    "count": 10,
-    "countLogical": 20
-  },
-  "gpu_nvidia": [
-    {
-      "name": "NVIDIA GeForce RTX 4070 Ti",
-      "memoryTotal": "12878610432",
-      "cudaCores": 7680,
-      "architecture": "Ada"
-    },
-    {
-      "name": "NVIDIA GeForce RTX 4070 Ti",
-      "memoryTotal": "12878610432",
-      "cudaCores": 7680,
-      "architecture": "Ada"
-    }
-  ],
-  "cudaVersion": "13.0"
-}

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/files/wandb-summary.json

@@ -1 +0,0 @@
-{"global_step":17486,"loss/q2":43288.23046875,"train/mean_score":-1.2233818703918429,"_runtime":65171.13856882,"train_loss":50942.46326462976,"loss/q1":43442.93359375,"loss/v":584.6388549804688,"stats/done_mean":0.018518518656492233,"test/cumulative_reward":-100.61537228953237,"stats/bootstrap_mean":-411.8333435058594,"loss/v1":303.723388671875,"test/mean_score":-1.214229753178261,"_wandb":{"runtime":65171},"lr":0.0001,"loss/actor_kl":1.2818808555603027,"epoch":200,"loss/q":86731.1640625,"_step":17486,"stats/t_mean":49.76852035522461,"_timestamp":1.775456456854072e+09,"val_loss":54167.66971343214,"loss/v2":280.91546630859375,"train/cumulative_reward":-101.41718616077296,"train/success_rate":0,"test/success_rate":0,"stats/kl_mean":1.2818808555603027,"stats/reward_mean":0.48456788063049316,"loss/actor_bootstrap":-411.8333435058594,"loss/actor":413.1152038574219}

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug-core.log

@@ -1,14 +0,0 @@
-{"time":"2026-04-05T20:15:02.000521617+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmp4tsm4dtw/port-349699.txt","pid":349699,"debug":false,"disable-analytics":false}
-{"time":"2026-04-05T20:15:02.000551014+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
-{"time":"2026-04-05T20:15:02.004632764+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":41859,"Zone":""}}
-{"time":"2026-04-05T20:15:02.004700387+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":349699}
-{"time":"2026-04-05T20:15:02.197125426+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:45888"}
-{"time":"2026-04-05T20:15:02.674902878+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"gsxv5o2x","id":"127.0.0.1:45888"}
-{"time":"2026-04-05T20:15:02.783690422+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"gsxv5o2x","id":"127.0.0.1:45888"}
-{"time":"2026-04-06T14:21:13.810553531+08:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:45888"}
-{"time":"2026-04-06T14:21:13.810680085+08:00","level":"INFO","msg":"server is shutting down"}
-{"time":"2026-04-06T14:21:13.810660834+08:00","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:45888"}
-{"time":"2026-04-06T14:21:13.810854796+08:00","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:45888"}
-{"time":"2026-04-06T14:21:15.820439179+08:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:45888"}
-{"time":"2026-04-06T14:21:15.820478972+08:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:45888"}
-{"time":"2026-04-06T14:21:15.820503484+08:00","level":"INFO","msg":"server is closed"}

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug-internal.log

@@ -1,16 +0,0 @@
-{"time":"2026-04-05T20:15:02.675207147+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
-{"time":"2026-04-05T20:15:02.675236072+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug-core.log"}
-{"time":"2026-04-05T20:15:02.783639967+08:00","level":"INFO","msg":"created new stream","id":"gsxv5o2x"}
-{"time":"2026-04-05T20:15:02.783682009+08:00","level":"INFO","msg":"stream: started","id":"gsxv5o2x"}
-{"time":"2026-04-05T20:15:02.783719383+08:00","level":"INFO","msg":"sender: started","stream_id":"gsxv5o2x"}
-{"time":"2026-04-05T20:15:02.783704075+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-05T20:15:02.783701477+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-05T20:15:03.681306658+08:00","level":"INFO","msg":"Starting system monitor"}
-{"time":"2026-04-06T14:21:13.810659156+08:00","level":"INFO","msg":"stream: closing","id":"gsxv5o2x"}
-{"time":"2026-04-06T14:21:13.81070755+08:00","level":"INFO","msg":"Stopping system monitor"}
-{"time":"2026-04-06T14:21:13.811296491+08:00","level":"INFO","msg":"Stopped system monitor"}
-{"time":"2026-04-06T14:21:15.381579169+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
-{"time":"2026-04-06T14:21:15.820068161+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-06T14:21:15.820142874+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"gsxv5o2x"}}
-{"time":"2026-04-06T14:21:15.820175329+08:00","level":"INFO","msg":"sender: closed","stream_id":"gsxv5o2x"}
-{"time":"2026-04-06T14:21:15.820335911+08:00","level":"INFO","msg":"stream: closed","id":"gsxv5o2x"}

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug.log

@@ -1,27 +0,0 @@
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Configure stats pid to 349699
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
-2026-04-05 20:15:02,667 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/train.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./train.py'}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_setup.py:_flush():79] Applying login settings: {}
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug.log
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/logs/debug-internal.log
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:init():619] calling init triggers
-2026-04-05 20:15:02,668 INFO    MainThread:349699 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
-config: {'name': 'train_bdpo_unet_lowdim', '_target_': 'llmbc.workspace.train_bdpo_unet_lowdim_workspace.TrainBDPOUnetLowdimWorkspace', 'obs_dim': 46, 'action_dim': 26, 'task_name': 'adroit-hand-hammer-v1', 'exp_name': 'default', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'policy': {'_target_': 'llmbc.policy.bdpo_unet_lowdim_policy.BDPOUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 26, 'local_cond_dim': None, 'global_cond_dim': 46, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'obs_dim': 46, 'action_dim': 26, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'hidden_size': [256, 256], 'activation': 'relu', 'gamma': 0.99, 'tau': 0.005, 'eta': 1.0, 'rho': 1.0, 'target_reduction': 'min', 'max_q_backup': False, 'actor_update_interval': 1, 'value_warmup_steps': 1000, 'num_eval_sample_steps': 100, 'act_limit': None, 'squash_action': False, 'init_weight': 0.01, 'checkpoint_path': None, 'behavior_checkpoint_path': 'data/outputs/2026.04.04/12.01.08_train_diffusion_unet_lowdim_adroit-hand-hammer-v1/checkpoints/epoch=0095-val_loss=0.199.ckpt', 'init_actor_from_behavior': True}, 'dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 256, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'critic_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'value_optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.9, 0.999], 'eps': 1e-08, 'weight_decay': 0.0}, 'training': {'device': 'cuda:0', 'seed': 46, 'debug': False, 'resume': False, 'lr_scheduler': 'constant', 'lr_warmup_steps': 500, 'num_epochs': 201, 'gradient_accumulate_every': 1, 'grad_norm_clip': 5.0, 'rollout_every': 10, 'checkpoint_every': 10, 'val_every': 1, 'sample_every': 10, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'adroit-hand-hammer-v1-bdpo-training', 'resume': True, 'mode': 'online', 'name': '2026.04.05-20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'tags': ['train_bdpo_unet_lowdim', 'adroit-hand-hammer-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.04.05-20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}, 'task': {'name': 'adroit-hand-hammer-v1', 'obs_dim': 46, 'action_dim': 26, 'env_runner': {'_target_': 'llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner', 'env_name': 'llf-adroit-adroit-hand-hammer-v1', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 150, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset', 'data_path': 'datasets/adroit-hand-hammer-v1-general-reward-trans.pt', 'data_path2': 'datasets/adroit-hand-hammer-v1-reward-trans.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor'}}}
-2026-04-05 20:15:02,669 INFO    MainThread:349699 [wandb_init.py:init():669] starting backend
-2026-04-05 20:15:02,669 INFO    MainThread:349699 [wandb_init.py:init():673] sending inform_init request
-2026-04-05 20:15:02,671 INFO    MainThread:349699 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
-2026-04-05 20:15:02,671 INFO    MainThread:349699 [wandb_init.py:init():686] backend started and connected
-2026-04-05 20:15:02,678 INFO    MainThread:349699 [wandb_init.py:init():781] updated telemetry
-2026-04-05 20:15:02,741 INFO    MainThread:349699 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
-2026-04-05 20:15:03,675 INFO    MainThread:349699 [wandb_init.py:init():867] starting run threads in backend
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_console_start():2451] atexit reg
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_redirect():2299] redirect: wrap_raw
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_redirect():2364] Wrapping output streams.
-2026-04-05 20:15:04,113 INFO    MainThread:349699 [wandb_run.py:_redirect():2389] Redirects installed.
-2026-04-05 20:15:04,116 INFO    MainThread:349699 [wandb_init.py:init():911] run started, returning control to user process
-2026-04-05 20:15:04,117 INFO    MainThread:349699 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1'}
-2026-04-06 14:21:13,810 WARNING MsgRouterThr:349699 [router.py:message_loop():75] message_loop has been closed

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/run-20260405_201502-gsxv5o2x/run-gsxv5o2x.wandb

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:27db059839fc26b03a18c0800b0320f5557ef864b844cf691c21c91647888d78
-size 50732328

2026.04.05/20.14.49_train_bdpo_unet_lowdim_adroit-hand-hammer-v1/wandb/wandb-resume.json

@@ -1 +0,0 @@
-{"run_id": "gsxv5o2x"}