Upload folder using huggingface_hub

.gitattributes

@@ -42,3 +42,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 2026.01.26/11.48.33_train_llm_lowdim_box-close-v2/wandb/run-20260126_114842-zut8rsd7/run-zut8rsd7.wandb filter=lfs diff=lfs merge=lfs -text
 2026.01.26/11.49.03_train_llm_lowdim_box-close-v2/wandb/run-20260126_114912-eqoznxjg/run-eqoznxjg.wandb filter=lfs diff=lfs merge=lfs -text
 2026.01.26/12.30.16_train_llm_lowdim_box-close-v2/wandb/run-20260126_123023-egujxdao/run-egujxdao.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/run-raachjxu.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/run-raachjxu.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/run-ccfd9xxl.wandb filter=lfs diff=lfs merge=lfs -text

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/.hydra/config.yaml

@@ -0,0 +1,163 @@
+name: train_llmbc_lowdim
+_target_: llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace
+obs_dim: ${task.obs_dim}
+action_dim: ${task.action_dim}
+task_name: ${task.name}
+exp_name: default
+model_name: ${llm.name}
+horizon: 1
+n_obs_steps: 1
+n_action_steps: 1
+n_latency_steps: 0
+past_action_visible: false
+llm_orig_expert_feedback: true
+llm_do_sample: false
+policy:
+  _target_: llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy
+  model:
+    _target_: llmbc.model.policy.policy_mlp.PolicyMLP
+    input_size: ${eval:'${n_obs_steps}*${obs_dim}'}
+    hidden_size:
+    - 256
+    - 256
+    output_size: ${eval:'${n_action_steps}*${action_dim}'}
+    activation: relu
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+  obs_dim: ${obs_dim}
+  action_dim: ${action_dim}
+  llm_discriminator:
+    _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+    task_id: ${task_name}
+    llm_translator:
+      _target_: llmbc.translator.llm_translator.LLMTranslator
+      cfg: ${llm}
+      obs_dim: ${task.obs_dim}
+      action_dim: ${task.action_dim}
+      horizon: ${horizon}
+      n_obs_steps: ${n_obs_steps}
+      n_action_steps: ${n_action_steps}
+  loss_bc_weight: 1.0
+  loss_llm_weight: 0.001
+  horizon: ${horizon}
+  n_obs_steps: ${n_obs_steps}
+  n_action_steps: ${n_action_steps}
+  normalize_llm_loss: true
+dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.AdamW
+  lr: 0.01
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda:0
+  seed: 42
+  debug: false
+  resume: false
+  lr_scheduler: cosine
+  lr_warmup_steps: 10
+  num_epochs: 20001
+  gradient_accumulate_every: 8
+  grad_norm_clip: 0.5
+  rollout_every: 100
+  checkpoint_every: 100
+  val_every: 1
+  sample_every: 100
+  sample_max_batch: 128
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: ${task.name}-training-split-ablation
+  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: box-close-v2
+  obs_dim: 9
+  action_dim: 4
+  env_runner:
+    _target_: llmbc.env_runner.metaworld_lowdim_runner.MetaworldLowdimRunner
+    env_name: llf-metaworld-box-close-v2
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 30
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    - fp
+    visual: false
+    discount: 0.9
+  dataset:
+    _target_: llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset
+    data_path: datasets/box-close-v2-o4-mini-split0.05.pt
+    data_path2: datasets/box-close-v2-o4-mini-split0.05.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: true
+  instructor:
+    _target_: llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor
+llm:
+  name: HuggingFaceTB/SmolLM2-135M-Instruct
+  model_name: SmolLM2-135M-Instruct
+  config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+  causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+  use_quantization: false
+  use_joint_mlp_projector: true
+  llm_mode: ete-finetuned
+  finetune_mode: orig
+  checkpoint: data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000
+  max_length: 100
+  lora_config:
+    r: 32
+    lora_alpha: 64
+    lora_dropout: 0.05
+    bias: none
+    task_type: CAUSAL_LM
+  prompter:
+    _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+    use_joint_mlp_projector: true
+  hydra:
+    job:
+      override_dirname: ${model_name}
+    run:
+      dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${model_name}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+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=42
+  job:
+    name: train
+    chdir: null
+    override_dirname: training.seed=42
+    id: ???
+    num: ???
+    config_name: llmbc_box-close-v2_o4-mini_split0.05.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /work/u1131674/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /work/u1131674/LLM-BC/config/data_efficiency
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2
+    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

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- training.seed=42

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=0100-test_success_rate=0.160.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22f161c891d32a16e1f25c104b2b9aa3a57c120633055ebb854b3ecd6dae61ec
+size 864584

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=0500-test_success_rate=0.780.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2812880df907aa389c47e00e965dc13dfc39c20829a359b733daae4e5d47c24e
+size 864584

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=0700-test_success_rate=0.760.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cf5ea1887ccf4fbe54ee0f62a3c42c8636858e5d7cab73a41db1886461230f1
+size 864584

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=1200-test_success_rate=0.760.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07061c3c037995b5755f7f10d60cd9842941c79ebe1b414fa338525f19827c0b
+size 864584

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=1600-test_success_rate=0.780.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7bf4a2477a2cac9ed1ad4be7517475fc7dc7d7581322acb3f22e003e7648cde
+size 864584

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=7900-test_success_rate=0.760.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70bdf629f036405c4028c1cdb1788aa9dc7af9c9798911176ed24681a70194c9
+size 864584

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/checkpoints/latest.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e32da9502909cec8e4929595d3503929356d81b482509099339493650f33ce3
+size 864584

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/logs.json.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c07fd38f5637dc4005b7cbb80a9b9f7811c967b19a896a276b2c5c328249d07
+size 55693005

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/train.log

@@ -0,0 +1,16 @@
+[2026-01-27 11:52:52,548][numexpr.utils][INFO] - Note: detected 112 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2026-01-27 11:52:52,548][numexpr.utils][INFO] - Note: NumExpr detected 112 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2026-01-27 11:52:52,549][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2026-01-27 11:52:52,558][numexpr.utils][INFO] - Note: detected 112 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2026-01-27 11:52:52,558][numexpr.utils][INFO] - Note: NumExpr detected 112 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2026-01-27 11:52:52,559][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2026-01-27 11:52:58,492][datasets][INFO] - PyTorch version 2.2.2 available.
+[2026-01-27 11:52:58,494][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2026-01-27 11:52:58,495][datasets][INFO] - JAX version 0.4.30 available.
+[2026-01-27 11:52:58,756][datasets][INFO] - PyTorch version 2.2.2 available.
+[2026-01-27 11:52:58,757][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2026-01-27 11:52:58,760][datasets][INFO] - JAX version 0.4.30 available.
+[2026-01-27 11:53:06,547][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
+[2026-01-27 11:53:06,558][absl][INFO] - MuJoCo library version is: 2.3.7
+[2026-01-27 11:53:07,104][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
+[2026-01-27 11:53:07,115][absl][INFO] - MuJoCo library version is: 2.3.7

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/debug-internal.log

@@ -0,0 +1,16 @@
+{"time":"2026-01-27T11:53:08.074868883+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-27T11:53:08.074878267+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug-core.log"}
+{"time":"2026-01-27T11:53:08.190280513+08:00","level":"INFO","msg":"created new stream","id":"raachjxu"}
+{"time":"2026-01-27T11:53:08.190312975+08:00","level":"INFO","msg":"stream: started","id":"raachjxu"}
+{"time":"2026-01-27T11:53:08.190338185+08:00","level":"INFO","msg":"sender: started","stream_id":"raachjxu"}
+{"time":"2026-01-27T11:53:08.190329168+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T11:53:08.1903328+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T11:53:08.732831596+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-01-27T22:37:36.57283309+08:00","level":"INFO","msg":"stream: closing","id":"raachjxu"}
+{"time":"2026-01-27T22:37:36.572875878+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-01-27T22:37:36.573489042+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-01-27T22:37:40.850119587+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-01-27T22:37:41.125467393+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T22:37:41.125498907+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T22:37:41.125540962+08:00","level":"INFO","msg":"sender: closed","stream_id":"raachjxu"}
+{"time":"2026-01-27T22:37:41.125562117+08:00","level":"INFO","msg":"stream: closed","id":"raachjxu"}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/debug.log

@@ -0,0 +1,28 @@
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Configure stats pid to 2419965
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/work/u1131674/LLM-BC/train.py', 'program': '/work/u1131674/LLM-BC/./train.py'}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug.log
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug-internal.log
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():619] calling init triggers
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llmbc_lowdim', '_target_': 'llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace', 'obs_dim': 9, 'action_dim': 4, 'task_name': 'box-close-v2', 'exp_name': 'default', 'model_name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy', 'model': {'_target_': 'llmbc.model.policy.policy_mlp.PolicyMLP', 'input_size': 9, 'hidden_size': [256, 256], 'output_size': 4, 'activation': 'relu', 'n_obs_steps': 1, 'n_action_steps': 1}, 'obs_dim': 9, 'action_dim': 4, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'box-close-v2', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_bc_weight': 1.0, 'loss_llm_weight': 0.001, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'normalize_llm_loss': True}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.01, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 43, 'debug': False, 'resume': False, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 20001, 'gradient_accumulate_every': 8, 'grad_norm_clip': 0.5, 'rollout_every': 100, 'checkpoint_every': 100, 'val_every': 1, 'sample_every': 100, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'box-close-v2-training-split-ablation', 'resume': True, 'mode': 'online', 'name': '2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2', 'tags': ['train_llmbc_lowdim', 'box-close-v2', '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/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2', 'wandb_name_base': '2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2'}, 'task': {'name': 'box-close-v2', 'obs_dim': 9, 'action_dim': 4, 'env_runner': {'_target_': 'llmbc.env_runner.metaworld_lowdim_runner.MetaworldLowdimRunner', 'env_name': 'llf-metaworld-box-close-v2', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 30, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.9}, 'dataset': {'_target_': 'llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset', 'data_path': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'data_path2': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': True}, 'instructor': {'_target_': 'llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():669] starting backend
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():673] sending inform_init request
+2026-01-27 11:53:08,072 INFO    MainThread:2419965 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-27 11:53:08,073 INFO    MainThread:2419965 [wandb_init.py:init():686] backend started and connected
+2026-01-27 11:53:08,082 INFO    MainThread:2419965 [wandb_init.py:init():781] updated telemetry
+2026-01-27 11:53:08,142 INFO    MainThread:2419965 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-27 11:53:08,717 INFO    MainThread:2419965 [wandb_init.py:init():859] run resumed
+2026-01-27 11:53:08,728 INFO    MainThread:2419965 [wandb_init.py:init():867] starting run threads in backend
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-27 11:53:09,294 INFO    MainThread:2419965 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-27 11:53:09,295 INFO    MainThread:2419965 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2'}
+2026-01-27 22:37:36,572 WARNING MsgRouterThr:2419965 [router.py:message_loop():75] message_loop has been closed

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/files/output.log

@@ -0,0 +1,3 @@
+Eval MetaworldLowdimRunner 1/6:   0%|          | 0/30 [00:00<?, ?it/s]/work/u1131674/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)
+                                                                                                    

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/files/requirements.txt

@@ -0,0 +1,858 @@
+rpds-py==0.27.1
+typeguard==4.4.4
+flatbuffers==25.12.19
+toppra==0.6.3
+sympy==1.14.0
+tiktoken==0.8.0
+nvidia-cuda-cupti-cu12==12.1.105
+arm_pytorch_utilities==0.4.3
+pynndescent==0.6.0
+multidict==6.7.0
+fonttools==4.60.2
+numexpr==2.10.1
+cmudict==1.0.13
+PyOpenGL-accelerate==3.1.10
+gmpy2==2.2.1
+peft==0.14.0
+metaworld==2.0.0
+nvidia-cufft-cu12==11.0.2.54
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+pexpect==4.9.0
+protobuf==4.25.8
+typing_extensions==4.15.0
+mujoco==2.3.7
+tokenizers==0.21.0
+pytorch-kinematics==0.7.5
+sniffio==1.3.1
+aiofiles==25.1.0
+mplib==0.1.1
+wcwidth==0.2.14
+Pygments==2.19.1
+anyio==4.12.1
+tensorflow-estimator==2.15.0
+filelock==3.17.0
+numpy==1.23.5
+attrs==25.4.0
+Markdown==3.9
+fsspec==2024.3.1
+libclang==18.1.1
+umap-learn==0.5.9.post2
+dill==0.3.8
+narwhals==2.15.0
+tensorboard==2.15.2
+dacite==1.9.2
+termcolor==3.1.0
+llmbc==0.0.0
+python-multipart==0.0.20
+exceptiongroup==1.3.1
+sapien==3.0.0b1
+pygame==2.6.1
+nvidia-curand-cu12==10.3.2.106
+evaluate==0.4.3
+msgpack==1.1.1
+tensorflow-probability==0.23.0
+diffusers==0.31.0
+certifi==2025.10.5
+d4rl==1.1
+pydub==0.25.1
+annotated-doc==0.0.4
+gitdb==4.0.12
+gradio_client==0.2.9
+Shapely==1.8.4
+mani_skill==3.0.0b20
+tensorflow-io-gcs-filesystem==0.37.1
+fasteners==0.20
+hjson==3.1.0
+ninja==1.13.0
+stack-data==0.6.3
+pyarrow==21.0.0
+networkx==3.2.1
+nvidia-cusparse-cu12==12.1.0.106
+pyparsing==3.3.1
+timm==1.0.22
+typing-inspection==0.4.2
+openai==2.8.1
+pybullet==3.2.6
+hydra-core==1.2.0
+gradio==3.36.1
+tensorflow==2.15.1
+asttokens==3.0.1
+importlib-metadata==5.2.0
+astunparse==1.6.3
+tifffile==2024.8.30
+annotated-types==0.7.0
+Bottleneck==1.4.2
+accelerate==1.0.1
+pytz==2025.2
+urllib3==2.5.0
+frozenlist==1.8.0
+sentry-sdk==2.50.0
+jsonschema==4.25.1
+tyro==0.9.1
+Farama-Notifications==0.0.4
+ffmpy==1.0.0
+httpx==0.28.1
+pymunk==6.2.1
+shtab==1.7.2
+glfw==2.0.0
+hf-xet==1.1.8
+omegaconf==2.2.1
+blobfile==3.0.0
+decorator==5.2.1
+cffi==1.17.1
+matplotlib-inline==0.2.1
+eval_type_backport==0.2.2
+torchaudio==2.2.2
+colorama==0.4.6
+click==8.1.8
+Cython==0.29.37
+orjson==3.11.5
+gym_bandits==0.0.2
+traitlets==5.14.3
+docker-pycreds==0.4.0
+multiprocess==0.70.15
+zipp==3.21.0
+antlr4-python3-runtime==4.9.3
+uc-micro-py==1.0.3
+mpmath==1.3.0
+idna==3.11
+aiodns==3.5.0
+charset-normalizer==3.4.4
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-cuda-nvrtc-cu12==12.1.105
+seaborn==0.13.2
+pyarrow-hotfix==0.7
+pillow==11.3.0
+pyautogen==0.1.0
+requests==2.32.0
+MarkupSafe==3.0.2
+websockets==15.0.1
+nvidia-nccl-cu12==2.19.3
+pure_eval==0.2.3
+parso==0.8.5
+huggingface-hub==0.26.2
+syllables==1.0.9
+tf-agents==0.19.0
+six==1.17.0
+referencing==0.36.2
+ptyprocess==0.7.0
+platformdirs==4.4.0
+fastapi==0.128.0
+stable-baselines3==2.2.1
+av==10.0.0
+diskcache==5.6.3
+pynvml==13.0.1
+pytorch-seed==0.2.0
+zarr==2.12.0
+mdurl==0.1.2
+docstring-parser==0.16
+packaging==25.0
+numcodecs==0.12.1
+opt_einsum==3.4.0
+markdown-it-py==2.2.0
+nvidia-cuda-runtime-cu12==12.1.105
+PyWavelets==1.6.0
+datasets==2.19.0
+contourpy==1.3.0
+aiohappyeyeballs==2.6.1
+jaxlib==0.4.30
+ImageIO==2.37.2
+wandb==0.18.6
+jiter==0.12.0
+gymnasium==0.29.1
+pycryptodomex==3.23.0
+google-pasta==0.2.0
+ipython==8.18.1
+threadpoolctl==3.6.0
+py-cpuinfo==9.0.0
+bitsandbytes==0.45.0
+xxhash==3.5.0
+google-auth-oauthlib==1.2.4
+rsa==4.9.1
+rouge_score==0.1.2
+dm-control==1.0.14
+oauthlib==3.3.1
+pandas==2.3.3
+tenacity==9.1.2
+asciitree==0.3.3
+scipy==1.13.1
+jedi==0.19.2
+gast==0.7.0
+google-auth==2.47.0
+transforms3d==0.4.2
+kiwisolver==1.4.7
+matplotlib==3.7.5
+aiohttp==3.12.15
+pip==23.3.2
+imageio-ffmpeg==0.6.0
+deepspeed==0.16.1
+yarl==1.18.0
+nvidia-nvtx-cu12==12.1.105
+llfbench==0.1.0
+wheel==0.45.1
+PySocks==1.7.1
+ml-dtypes==0.3.2
+PyYAML==6.0.2
+fast_kinematics==0.2.2
+gin-config==0.5.0
+setproctitle==1.3.7
+safetensors==0.5.3
+torchvision==0.17.2
+semantic-version==2.10.0
+PyOpenGL==3.1.10
+nltk==3.9.2
+lxml==6.0.2
+pydantic==2.12.5
+tqdm==4.67.1
+keras==2.15.0
+parse==1.19.1
+linkify-it-py==2.0.3
+dm-tree==0.1.8
+requests-oauthlib==2.0.0
+scikit-learn==1.6.1
+altair==6.0.0
+Werkzeug==3.1.5
+sentencepiece==0.2.0
+uvicorn==0.39.0
+cycler==0.12.1
+transformers==4.47.1
+uvloop==0.22.1
+mkl_random==1.2.8
+GitPython==3.1.46
+regex==2025.9.1
+jax==0.4.30
+llvmlite==0.39.1
+pyasn1_modules==0.4.2
+nvidia-cudnn-cu12==8.9.2.26
+pydantic_core==2.41.5
+google-genai==1.47.0
+propcache==0.3.1
+pycares==4.10.0
+pyperclip==1.11.0
+pyasn1==0.6.2
+async-timeout==5.0.1
+psutil==7.0.0
+gym==0.23.1
+dm-env==1.6
+Jinja2==3.1.6
+sentence-transformers==3.2.1
+einops==0.4.1
+triton==2.2.0
+grpcio==1.76.0
+labmaze==1.0.6
+nvidia-ml-py==13.590.44
+brotlicffi==1.0.9.2
+smmap==5.0.2
+cloudpickle==3.1.2
+setuptools==80.9.0
+starlette==0.49.3
+prompt_toolkit==3.0.52
+wrapt==1.14.2
+h5py==3.14.0
+scikit-image==0.19.3
+joblib==1.5.3
+opencv-python==4.11.0.86
+rich==14.2.0
+trl==0.11.4
+gym-notices==0.1.0
+trimesh==4.11.1
+mdit-py-plugins==0.3.3
+distro==1.9.0
+executing==2.2.1
+mkl-service==2.4.0
+nvidia-cusolver-cu12==11.4.5.107
+FLAML==2.3.6
+mujoco-py==2.1.2.14
+h11==0.16.0
+highway-env==1.9.1
+httpcore==1.0.9
+tensorboard-data-server==0.7.2
+tzdata==2025.3
+absl-py==2.3.1
+jsonschema-specifications==2025.9.1
+numba==0.56.4
+tabulate==0.9.0
+importlib-resources==5.13.0
+pycparser==2.23
+mkl_fft==1.3.11
+torch==2.2.2
+nvidia-cublas-cu12==12.1.3.1
+rpds-py==0.27.1
+typeguard==4.4.4
+flatbuffers==25.12.19
+toppra==0.6.3
+sympy==1.14.0
+tiktoken==0.8.0
+nvidia-cuda-cupti-cu12==12.1.105
+arm_pytorch_utilities==0.4.3
+pynndescent==0.6.0
+multidict==6.7.0
+fonttools==4.60.2
+numexpr==2.10.1
+cmudict==1.0.13
+PyOpenGL-accelerate==3.1.10
+gmpy2==2.2.1
+peft==0.14.0
+metaworld==2.0.0
+nvidia-cufft-cu12==11.0.2.54
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+pexpect==4.9.0
+protobuf==4.25.8
+typing_extensions==4.15.0
+mujoco==2.3.7
+tokenizers==0.21.0
+pytorch-kinematics==0.7.5
+sniffio==1.3.1
+aiofiles==25.1.0
+mplib==0.1.1
+wcwidth==0.2.14
+Pygments==2.19.1
+anyio==4.12.1
+tensorflow-estimator==2.15.0
+filelock==3.17.0
+numpy==1.23.5
+attrs==25.4.0
+Markdown==3.9
+fsspec==2024.3.1
+libclang==18.1.1
+umap-learn==0.5.9.post2
+dill==0.3.8
+narwhals==2.15.0
+tensorboard==2.15.2
+dacite==1.9.2
+termcolor==3.1.0
+llmbc==0.0.0
+python-multipart==0.0.20
+exceptiongroup==1.3.1
+sapien==3.0.0b1
+pygame==2.6.1
+nvidia-curand-cu12==10.3.2.106
+evaluate==0.4.3
+msgpack==1.1.1
+tensorflow-probability==0.23.0
+diffusers==0.31.0
+certifi==2025.10.5
+d4rl==1.1
+pydub==0.25.1
+annotated-doc==0.0.4
+gitdb==4.0.12
+gradio_client==0.2.9
+Shapely==1.8.4
+mani_skill==3.0.0b20
+tensorflow-io-gcs-filesystem==0.37.1
+fasteners==0.20
+hjson==3.1.0
+ninja==1.13.0
+stack-data==0.6.3
+pyarrow==21.0.0
+networkx==3.2.1
+nvidia-cusparse-cu12==12.1.0.106
+pyparsing==3.3.1
+timm==1.0.22
+typing-inspection==0.4.2
+openai==2.8.1
+pybullet==3.2.6
+hydra-core==1.2.0
+gradio==3.36.1
+tensorflow==2.15.1
+asttokens==3.0.1
+importlib-metadata==5.2.0
+astunparse==1.6.3
+tifffile==2024.8.30
+annotated-types==0.7.0
+Bottleneck==1.4.2
+accelerate==1.0.1
+pytz==2025.2
+urllib3==2.5.0
+frozenlist==1.8.0
+sentry-sdk==2.50.0
+jsonschema==4.25.1
+tyro==0.9.1
+Farama-Notifications==0.0.4
+ffmpy==1.0.0
+httpx==0.28.1
+pymunk==6.2.1
+shtab==1.7.2
+glfw==2.0.0
+hf-xet==1.1.8
+omegaconf==2.2.1
+blobfile==3.0.0
+decorator==5.2.1
+cffi==1.17.1
+matplotlib-inline==0.2.1
+eval_type_backport==0.2.2
+torchaudio==2.2.2
+colorama==0.4.6
+click==8.1.8
+Cython==0.29.37
+orjson==3.11.5
+gym_bandits==0.0.2
+traitlets==5.14.3
+docker-pycreds==0.4.0
+multiprocess==0.70.15
+zipp==3.21.0
+antlr4-python3-runtime==4.9.3
+uc-micro-py==1.0.3
+mpmath==1.3.0
+idna==3.11
+aiodns==3.5.0
+charset-normalizer==3.4.4
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-cuda-nvrtc-cu12==12.1.105
+seaborn==0.13.2
+pyarrow-hotfix==0.7
+pillow==11.3.0
+pyautogen==0.1.0
+requests==2.32.0
+MarkupSafe==3.0.2
+websockets==15.0.1
+nvidia-nccl-cu12==2.19.3
+pure_eval==0.2.3
+parso==0.8.5
+huggingface-hub==0.26.2
+syllables==1.0.9
+tf-agents==0.19.0
+six==1.17.0
+referencing==0.36.2
+ptyprocess==0.7.0
+platformdirs==4.4.0
+fastapi==0.128.0
+stable-baselines3==2.2.1
+av==10.0.0
+diskcache==5.6.3
+pynvml==13.0.1
+pytorch-seed==0.2.0
+zarr==2.12.0
+mdurl==0.1.2
+docstring-parser==0.16
+packaging==25.0
+numcodecs==0.12.1
+opt_einsum==3.4.0
+markdown-it-py==2.2.0
+nvidia-cuda-runtime-cu12==12.1.105
+PyWavelets==1.6.0
+datasets==2.19.0
+contourpy==1.3.0
+aiohappyeyeballs==2.6.1
+jaxlib==0.4.30
+ImageIO==2.37.2
+wandb==0.18.6
+jiter==0.12.0
+gymnasium==0.29.1
+pycryptodomex==3.23.0
+google-pasta==0.2.0
+ipython==8.18.1
+threadpoolctl==3.6.0
+py-cpuinfo==9.0.0
+bitsandbytes==0.45.0
+xxhash==3.5.0
+google-auth-oauthlib==1.2.4
+rsa==4.9.1
+rouge_score==0.1.2
+dm-control==1.0.14
+oauthlib==3.3.1
+pandas==2.3.3
+tenacity==9.1.2
+asciitree==0.3.3
+scipy==1.13.1
+jedi==0.19.2
+gast==0.7.0
+google-auth==2.47.0
+transforms3d==0.4.2
+kiwisolver==1.4.7
+matplotlib==3.7.5
+aiohttp==3.12.15
+pip==23.3.2
+imageio-ffmpeg==0.6.0
+deepspeed==0.16.1
+yarl==1.18.0
+nvidia-nvtx-cu12==12.1.105
+llfbench==0.1.0
+wheel==0.45.1
+PySocks==1.7.1
+ml-dtypes==0.3.2
+PyYAML==6.0.2
+fast_kinematics==0.2.2
+gin-config==0.5.0
+setproctitle==1.3.7
+safetensors==0.5.3
+torchvision==0.17.2
+semantic-version==2.10.0
+PyOpenGL==3.1.10
+nltk==3.9.2
+lxml==6.0.2
+pydantic==2.12.5
+tqdm==4.67.1
+keras==2.15.0
+parse==1.19.1
+linkify-it-py==2.0.3
+dm-tree==0.1.8
+requests-oauthlib==2.0.0
+scikit-learn==1.6.1
+altair==6.0.0
+Werkzeug==3.1.5
+sentencepiece==0.2.0
+uvicorn==0.39.0
+cycler==0.12.1
+transformers==4.47.1
+uvloop==0.22.1
+mkl_random==1.2.8
+GitPython==3.1.46
+regex==2025.9.1
+jax==0.4.30
+llvmlite==0.39.1
+pyasn1_modules==0.4.2
+nvidia-cudnn-cu12==8.9.2.26
+pydantic_core==2.41.5
+google-genai==1.47.0
+propcache==0.3.1
+pycares==4.10.0
+pyperclip==1.11.0
+pyasn1==0.6.2
+async-timeout==5.0.1
+psutil==7.0.0
+gym==0.23.1
+dm-env==1.6
+Jinja2==3.1.6
+sentence-transformers==3.2.1
+einops==0.4.1
+triton==2.2.0
+grpcio==1.76.0
+labmaze==1.0.6
+nvidia-ml-py==13.590.44
+brotlicffi==1.0.9.2
+smmap==5.0.2
+cloudpickle==3.1.2
+setuptools==80.9.0
+starlette==0.49.3
+prompt_toolkit==3.0.52
+wrapt==1.14.2
+h5py==3.14.0
+scikit-image==0.19.3
+joblib==1.5.3
+opencv-python==4.11.0.86
+rich==14.2.0
+trl==0.11.4
+gym-notices==0.1.0
+trimesh==4.11.1
+mdit-py-plugins==0.3.3
+distro==1.9.0
+executing==2.2.1
+mkl-service==2.4.0
+nvidia-cusolver-cu12==11.4.5.107
+FLAML==2.3.6
+mujoco-py==2.1.2.14
+h11==0.16.0
+highway-env==1.9.1
+httpcore==1.0.9
+tensorboard-data-server==0.7.2
+tzdata==2025.3
+absl-py==2.3.1
+jsonschema-specifications==2025.9.1
+numba==0.56.4
+tabulate==0.9.0
+importlib-resources==5.13.0
+pycparser==2.23
+mkl_fft==1.3.11
+torch==2.2.2
+nvidia-cublas-cu12==12.1.3.1
+llmbc==0.0.0
+llmbc==0.0.0
+rpds-py==0.27.1
+typeguard==4.4.4
+flatbuffers==25.12.19
+toppra==0.6.3
+sympy==1.14.0
+tiktoken==0.8.0
+nvidia-cuda-cupti-cu12==12.1.105
+arm_pytorch_utilities==0.4.3
+pynndescent==0.6.0
+multidict==6.7.0
+fonttools==4.60.2
+numexpr==2.10.1
+cmudict==1.0.13
+PyOpenGL-accelerate==3.1.10
+gmpy2==2.2.1
+peft==0.14.0
+metaworld==2.0.0
+nvidia-cufft-cu12==11.0.2.54
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+pexpect==4.9.0
+protobuf==4.25.8
+typing_extensions==4.15.0
+mujoco==2.3.7
+tokenizers==0.21.0
+pytorch-kinematics==0.7.5
+sniffio==1.3.1
+aiofiles==25.1.0
+mplib==0.1.1
+wcwidth==0.2.14
+Pygments==2.19.1
+anyio==4.12.1
+tensorflow-estimator==2.15.0
+filelock==3.17.0
+numpy==1.23.5
+attrs==25.4.0
+Markdown==3.9
+fsspec==2024.3.1
+libclang==18.1.1
+umap-learn==0.5.9.post2
+dill==0.3.8
+narwhals==2.15.0
+tensorboard==2.15.2
+dacite==1.9.2
+termcolor==3.1.0
+llmbc==0.0.0
+python-multipart==0.0.20
+exceptiongroup==1.3.1
+sapien==3.0.0b1
+pygame==2.6.1
+nvidia-curand-cu12==10.3.2.106
+evaluate==0.4.3
+msgpack==1.1.1
+tensorflow-probability==0.23.0
+diffusers==0.31.0
+certifi==2025.10.5
+d4rl==1.1
+pydub==0.25.1
+annotated-doc==0.0.4
+gitdb==4.0.12
+gradio_client==0.2.9
+Shapely==1.8.4
+mani_skill==3.0.0b20
+tensorflow-io-gcs-filesystem==0.37.1
+fasteners==0.20
+hjson==3.1.0
+ninja==1.13.0
+stack-data==0.6.3
+pyarrow==21.0.0
+networkx==3.2.1
+nvidia-cusparse-cu12==12.1.0.106
+pyparsing==3.3.1
+timm==1.0.22
+typing-inspection==0.4.2
+openai==2.8.1
+pybullet==3.2.6
+hydra-core==1.2.0
+gradio==3.36.1
+tensorflow==2.15.1
+asttokens==3.0.1
+importlib-metadata==5.2.0
+astunparse==1.6.3
+tifffile==2024.8.30
+annotated-types==0.7.0
+Bottleneck==1.4.2
+accelerate==1.0.1
+pytz==2025.2
+urllib3==2.5.0
+frozenlist==1.8.0
+sentry-sdk==2.50.0
+jsonschema==4.25.1
+tyro==0.9.1
+Farama-Notifications==0.0.4
+ffmpy==1.0.0
+httpx==0.28.1
+pymunk==6.2.1
+shtab==1.7.2
+glfw==2.0.0
+hf-xet==1.1.8
+omegaconf==2.2.1
+blobfile==3.0.0
+decorator==5.2.1
+cffi==1.17.1
+matplotlib-inline==0.2.1
+eval_type_backport==0.2.2
+torchaudio==2.2.2
+colorama==0.4.6
+click==8.1.8
+Cython==0.29.37
+orjson==3.11.5
+gym_bandits==0.0.2
+traitlets==5.14.3
+docker-pycreds==0.4.0
+multiprocess==0.70.15
+zipp==3.21.0
+antlr4-python3-runtime==4.9.3
+uc-micro-py==1.0.3
+mpmath==1.3.0
+idna==3.11
+aiodns==3.5.0
+charset-normalizer==3.4.4
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-cuda-nvrtc-cu12==12.1.105
+seaborn==0.13.2
+pyarrow-hotfix==0.7
+pillow==11.3.0
+pyautogen==0.1.0
+requests==2.32.0
+MarkupSafe==3.0.2
+websockets==15.0.1
+nvidia-nccl-cu12==2.19.3
+pure_eval==0.2.3
+parso==0.8.5
+huggingface-hub==0.26.2
+syllables==1.0.9
+tf-agents==0.19.0
+six==1.17.0
+referencing==0.36.2
+ptyprocess==0.7.0
+platformdirs==4.4.0
+fastapi==0.128.0
+stable-baselines3==2.2.1
+av==10.0.0
+diskcache==5.6.3
+pynvml==13.0.1
+pytorch-seed==0.2.0
+zarr==2.12.0
+mdurl==0.1.2
+docstring-parser==0.16
+packaging==25.0
+numcodecs==0.12.1
+opt_einsum==3.4.0
+markdown-it-py==2.2.0
+nvidia-cuda-runtime-cu12==12.1.105
+PyWavelets==1.6.0
+datasets==2.19.0
+contourpy==1.3.0
+aiohappyeyeballs==2.6.1
+jaxlib==0.4.30
+ImageIO==2.37.2
+wandb==0.18.6
+jiter==0.12.0
+gymnasium==0.29.1
+pycryptodomex==3.23.0
+google-pasta==0.2.0
+ipython==8.18.1
+threadpoolctl==3.6.0
+py-cpuinfo==9.0.0
+bitsandbytes==0.45.0
+xxhash==3.5.0
+google-auth-oauthlib==1.2.4
+rsa==4.9.1
+rouge_score==0.1.2
+dm-control==1.0.14
+oauthlib==3.3.1
+pandas==2.3.3
+tenacity==9.1.2
+asciitree==0.3.3
+scipy==1.13.1
+jedi==0.19.2
+gast==0.7.0
+google-auth==2.47.0
+transforms3d==0.4.2
+kiwisolver==1.4.7
+matplotlib==3.7.5
+aiohttp==3.12.15
+pip==23.3.2
+imageio-ffmpeg==0.6.0
+deepspeed==0.16.1
+yarl==1.18.0
+nvidia-nvtx-cu12==12.1.105
+llfbench==0.1.0
+wheel==0.45.1
+PySocks==1.7.1
+ml-dtypes==0.3.2
+PyYAML==6.0.2
+fast_kinematics==0.2.2
+gin-config==0.5.0
+setproctitle==1.3.7
+safetensors==0.5.3
+torchvision==0.17.2
+semantic-version==2.10.0
+PyOpenGL==3.1.10
+nltk==3.9.2
+lxml==6.0.2
+pydantic==2.12.5
+tqdm==4.67.1
+keras==2.15.0
+parse==1.19.1
+linkify-it-py==2.0.3
+dm-tree==0.1.8
+requests-oauthlib==2.0.0
+scikit-learn==1.6.1
+altair==6.0.0
+Werkzeug==3.1.5
+sentencepiece==0.2.0
+uvicorn==0.39.0
+cycler==0.12.1
+transformers==4.47.1
+uvloop==0.22.1
+mkl_random==1.2.8
+GitPython==3.1.46
+regex==2025.9.1
+jax==0.4.30
+llvmlite==0.39.1
+pyasn1_modules==0.4.2
+nvidia-cudnn-cu12==8.9.2.26
+pydantic_core==2.41.5
+google-genai==1.47.0
+propcache==0.3.1
+pycares==4.10.0
+pyperclip==1.11.0
+pyasn1==0.6.2
+async-timeout==5.0.1
+psutil==7.0.0
+gym==0.23.1
+dm-env==1.6
+Jinja2==3.1.6
+sentence-transformers==3.2.1
+einops==0.4.1
+triton==2.2.0
+grpcio==1.76.0
+labmaze==1.0.6
+nvidia-ml-py==13.590.44
+brotlicffi==1.0.9.2
+smmap==5.0.2
+cloudpickle==3.1.2
+setuptools==80.9.0
+starlette==0.49.3
+prompt_toolkit==3.0.52
+wrapt==1.14.2
+h5py==3.14.0
+scikit-image==0.19.3
+joblib==1.5.3
+opencv-python==4.11.0.86
+rich==14.2.0
+trl==0.11.4
+gym-notices==0.1.0
+trimesh==4.11.1
+mdit-py-plugins==0.3.3
+distro==1.9.0
+executing==2.2.1
+mkl-service==2.4.0
+nvidia-cusolver-cu12==11.4.5.107
+FLAML==2.3.6
+mujoco-py==2.1.2.14
+h11==0.16.0
+highway-env==1.9.1
+httpcore==1.0.9
+tensorboard-data-server==0.7.2
+tzdata==2025.3
+absl-py==2.3.1
+jsonschema-specifications==2025.9.1
+numba==0.56.4
+tabulate==0.9.0
+importlib-resources==5.13.0
+pycparser==2.23
+mkl_fft==1.3.11
+torch==2.2.2
+nvidia-cublas-cu12==12.1.3.1
+zipp==3.19.2
+jaraco.text==3.12.1
+jaraco.context==5.3.0
+importlib_metadata==8.0.0
+typeguard==4.3.0
+inflect==7.3.1
+more-itertools==10.3.0
+wheel==0.45.1
+packaging==24.2
+backports.tarfile==1.2.0
+autocommand==2.2.2
+jaraco.collections==5.1.0
+tomli==2.0.1
+platformdirs==4.2.2
+jaraco.functools==4.0.1
+typing_extensions==4.12.2

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/files/wandb-metadata.json

@@ -0,0 +1,107 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.28",
+  "python": "3.9.25",
+  "startedAt": "2026-01-27T03:53:07.506395Z",
+  "args": [
+    "--config-path",
+    "config/data_efficiency",
+    "--config-name",
+    "llmbc_box-close-v2_o4-mini_split0.05.yaml",
+    "training.seed=42"
+  ],
+  "program": "/work/u1131674/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "206b9c57fa687707dfa1ee0ebb33b507135a59c6"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2",
+  "host": "hgpn05",
+  "username": "u1131674",
+  "executable": "/home/u1131674/.conda/envs/llm-bc/bin/python3",
+  "codePathLocal": "train.py",
+  "cpu_count": 112,
+  "cpu_count_logical": 112,
+  "gpu": "NVIDIA H100 80GB HBM3",
+  "gpu_count": 1,
+  "disk": {
+    "/": {
+      "total": "1918024196096",
+      "used": "437322797056"
+    }
+  },
+  "memory": {
+    "total": "2163617771520"
+  },
+  "cpu": {
+    "count": 112,
+    "countLogical": 112
+  },
+  "gpu_nvidia": [
+    {
+      "name": "NVIDIA H100 80GB HBM3",
+      "memoryTotal": "85520809984",
+      "cudaCores": 16896,
+      "architecture": "Hopper"
+    }
+  ],
+  "slurm": {
+    "cluster_name": "hpc",
+    "conf": "/etc/slurm/slurm.conf",
+    "cpu_bind": "quiet,mask_cpu:0x000000000FE04000000000000000",
+    "cpu_bind_list": "0x000000000FE04000000000000000",
+    "cpu_bind_type": "mask_cpu:",
+    "cpu_bind_verbose": "quiet",
+    "cpus_on_node": "8",
+    "cpus_per_task": "8",
+    "distribution": "cyclic,pack",
+    "gpus_on_node": "1",
+    "gpus_per_node": "1",
+    "gtids": "0",
+    "job_account": "mst114558",
+    "job_cpus_per_node": "8",
+    "job_end_time": "1769658768",
+    "job_gid": "106773",
+    "job_group": "MST114558",
+    "job_id": "104242",
+    "job_name": "python3",
+    "job_nodelist": "hgpn05",
+    "job_num_nodes": "1",
+    "job_partition": "normal",
+    "job_qos": "normal",
+    "job_start_time": "1769485968",
+    "job_uid": "41408",
+    "job_user": "u1131674",
+    "jobid": "104242",
+    "launch_node_ipaddr": "172.21.101.1",
+    "localid": "0",
+    "mem_per_node": "204800",
+    "nnodes": "1",
+    "nodeid": "0",
+    "nodelist": "hgpn05",
+    "nprocs": "1",
+    "ntasks": "1",
+    "prio_process": "0",
+    "procid": "0",
+    "srun_comm_host": "172.21.101.1",
+    "srun_comm_port": "42069",
+    "step_gpus": "5",
+    "step_id": "0",
+    "step_launcher_port": "42069",
+    "step_nodelist": "hgpn05",
+    "step_num_nodes": "1",
+    "step_num_tasks": "1",
+    "step_tasks_per_node": "1",
+    "stepid": "0",
+    "submit_dir": "/work/u1131674/LLM-BC",
+    "submit_host": "cbi-lgn01",
+    "task_pid": "754949",
+    "tasks_per_node": "1",
+    "topology_addr": "ibsw1.hgpn05",
+    "topology_addr_pattern": "switch.node",
+    "tres_per_task": "cpu:8",
+    "umask": "0022"
+  },
+  "cudaVersion": "12.4"
+}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/logs/debug-core.log

@@ -0,0 +1,7 @@
+{"time":"2026-01-27T11:53:06.897507519+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpi1osecmd/port-754949.txt","pid":754949,"debug":false,"disable-analytics":false}
+{"time":"2026-01-27T11:53:06.897527345+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-27T11:53:06.897861154+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":754949}
+{"time":"2026-01-27T11:53:06.897850599+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":32979,"Zone":""}}
+{"time":"2026-01-27T11:53:07.091367358+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:53454"}
+{"time":"2026-01-27T11:53:07.506927211+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"raachjxu","id":"127.0.0.1:53454"}
+{"time":"2026-01-27T11:53:07.624448731+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"raachjxu","id":"127.0.0.1:53454"}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/logs/debug-internal.log

@@ -0,0 +1,8 @@
+{"time":"2026-01-27T11:53:07.507948839+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-27T11:53:07.507957394+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/logs/debug-core.log"}
+{"time":"2026-01-27T11:53:07.624409739+08:00","level":"INFO","msg":"created new stream","id":"raachjxu"}
+{"time":"2026-01-27T11:53:07.624443283+08:00","level":"INFO","msg":"stream: started","id":"raachjxu"}
+{"time":"2026-01-27T11:53:07.624479516+08:00","level":"INFO","msg":"sender: started","stream_id":"raachjxu"}
+{"time":"2026-01-27T11:53:07.624459006+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T11:53:07.624475123+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T11:53:08.339033565+08:00","level":"INFO","msg":"Starting system monitor"}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/logs/debug.log

@@ -0,0 +1,26 @@
+2026-01-27 11:53:07,503 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Configure stats pid to 754949
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/work/u1131674/LLM-BC/train.py', 'program': '/work/u1131674/LLM-BC/./train.py'}
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/logs/debug.log
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/logs/debug-internal.log
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_init.py:init():619] calling init triggers
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llmbc_lowdim', '_target_': 'llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace', 'obs_dim': 9, 'action_dim': 4, 'task_name': 'box-close-v2', 'exp_name': 'default', 'model_name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy', 'model': {'_target_': 'llmbc.model.policy.policy_mlp.PolicyMLP', 'input_size': 9, 'hidden_size': [256, 256], 'output_size': 4, 'activation': 'relu', 'n_obs_steps': 1, 'n_action_steps': 1}, 'obs_dim': 9, 'action_dim': 4, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'box-close-v2', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_bc_weight': 1.0, 'loss_llm_weight': 0.001, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'normalize_llm_loss': True}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.01, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 42, 'debug': False, 'resume': False, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 20001, 'gradient_accumulate_every': 8, 'grad_norm_clip': 0.5, 'rollout_every': 100, 'checkpoint_every': 100, 'val_every': 1, 'sample_every': 100, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'box-close-v2-training-split-ablation', 'resume': True, 'mode': 'online', 'name': '2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2', 'tags': ['train_llmbc_lowdim', 'box-close-v2', '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/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2', 'wandb_name_base': '2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2'}, 'task': {'name': 'box-close-v2', 'obs_dim': 9, 'action_dim': 4, 'env_runner': {'_target_': 'llmbc.env_runner.metaworld_lowdim_runner.MetaworldLowdimRunner', 'env_name': 'llf-metaworld-box-close-v2', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 30, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.9}, 'dataset': {'_target_': 'llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset', 'data_path': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'data_path2': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': True}, 'instructor': {'_target_': 'llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_init.py:init():669] starting backend
+2026-01-27 11:53:07,504 INFO    MainThread:754949 [wandb_init.py:init():673] sending inform_init request
+2026-01-27 11:53:07,505 INFO    MainThread:754949 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-27 11:53:07,506 INFO    MainThread:754949 [wandb_init.py:init():686] backend started and connected
+2026-01-27 11:53:07,515 INFO    MainThread:754949 [wandb_init.py:init():781] updated telemetry
+2026-01-27 11:53:07,570 INFO    MainThread:754949 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-27 11:53:08,334 INFO    MainThread:754949 [wandb_init.py:init():867] starting run threads in backend
+2026-01-27 11:53:08,858 INFO    MainThread:754949 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-27 11:53:08,858 INFO    MainThread:754949 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-27 11:53:08,858 INFO    MainThread:754949 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-27 11:53:08,858 INFO    MainThread:754949 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-27 11:53:08,861 INFO    MainThread:754949 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-27 11:53:08,862 INFO    MainThread:754949 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2'}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115307-raachjxu/run-raachjxu.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77644bf1a49da807f34879bd1902e767c1fad49735292efcd9bb182c392eb3fb
+size 1769472

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/files/config.yaml

@@ -0,0 +1,272 @@
+_target_:
+    value: llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace
+_wandb:
+    value:
+        cli_version: 0.18.6
+        m: []
+        python_version: 3.9.25
+        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":
+                - 5
+                - 13
+                - 15
+                - 16
+                - 23
+                - 55
+                - 61
+            "4": 3.9.25
+            "5": 0.18.6
+            "6": 4.47.1
+            "8":
+                - 5
+            "12": 0.18.6
+            "13": linux-x86_64
+action_dim:
+    value: 4
+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
+dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: true
+exp_name:
+    value: default
+horizon:
+    value: 1
+llm:
+    value:
+        causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+        checkpoint: data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000
+        config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+        finetune_mode: orig
+        hydra:
+            job:
+                override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+            run:
+                dir: data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct
+        llm_mode: ete-finetuned
+        lora_config:
+            bias: none
+            lora_alpha: 64
+            lora_dropout: 0.05
+            r: 32
+            task_type: CAUSAL_LM
+        max_length: 100
+        model_name: SmolLM2-135M-Instruct
+        name: HuggingFaceTB/SmolLM2-135M-Instruct
+        prompter:
+            _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+            use_joint_mlp_projector: true
+        use_joint_mlp_projector: true
+        use_quantization: false
+llm_do_sample:
+    value: false
+llm_orig_expert_feedback:
+    value: true
+logging:
+    value:
+        group: null
+        id: null
+        mode: online
+        name: 2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2
+        project: box-close-v2-training-split-ablation
+        resume: true
+        tags:
+            - train_llmbc_lowdim
+            - box-close-v2
+            - default
+model_name:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+multi_run:
+    value:
+        run_dir: data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2
+        wandb_name_base: 2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2
+n_action_steps:
+    value: 1
+n_latency_steps:
+    value: 0
+n_obs_steps:
+    value: 1
+name:
+    value: train_llmbc_lowdim
+obs_dim:
+    value: 9
+optimizer:
+    value:
+        _target_: torch.optim.AdamW
+        betas:
+            - 0.95
+            - 0.999
+        eps: 1e-08
+        lr: 0.01
+        weight_decay: 1e-06
+output_dir:
+    value: /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2
+past_action_visible:
+    value: false
+policy:
+    value:
+        _target_: llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy
+        action_dim: 4
+        horizon: 1
+        llm_discriminator:
+            _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+            llm_translator:
+                _target_: llmbc.translator.llm_translator.LLMTranslator
+                action_dim: 4
+                cfg:
+                    causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+                    checkpoint: data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000
+                    config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+                    finetune_mode: orig
+                    hydra:
+                        job:
+                            override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+                        run:
+                            dir: data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct
+                    llm_mode: ete-finetuned
+                    lora_config:
+                        bias: none
+                        lora_alpha: 64
+                        lora_dropout: 0.05
+                        r: 32
+                        task_type: CAUSAL_LM
+                    max_length: 100
+                    model_name: SmolLM2-135M-Instruct
+                    name: HuggingFaceTB/SmolLM2-135M-Instruct
+                    prompter:
+                        _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+                        use_joint_mlp_projector: true
+                    use_joint_mlp_projector: true
+                    use_quantization: false
+                horizon: 1
+                n_action_steps: 1
+                n_obs_steps: 1
+                obs_dim: 9
+            task_id: box-close-v2
+        loss_bc_weight: 1
+        loss_llm_weight: 0.001
+        model:
+            _target_: llmbc.model.policy.policy_mlp.PolicyMLP
+            activation: relu
+            hidden_size:
+                - 256
+                - 256
+            input_size: 9
+            n_action_steps: 1
+            n_obs_steps: 1
+            output_size: 4
+        n_action_steps: 1
+        n_obs_steps: 1
+        normalize_llm_loss: true
+        obs_dim: 9
+task:
+    value:
+        action_dim: 4
+        dataset:
+            _target_: llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset
+            data_path: datasets/box-close-v2-o4-mini-split0.05.pt
+            data_path2: datasets/box-close-v2-o4-mini-split0.05.pt
+            dummy_normalizer: true
+            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.metaworld_lowdim_runner.MetaworldLowdimRunner
+            discount: 0.9
+            env_name: llf-metaworld-box-close-v2
+            feedback_type:
+                - hp
+                - hn
+                - fp
+            instruction_type: b
+            max_steps: 30
+            n_action_steps: 1
+            n_envs: 10
+            n_obs_steps: 1
+            n_test: 50
+            n_train: 10
+            visual: false
+        instructor:
+            _target_: llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor
+        name: box-close-v2
+        obs_dim: 9
+task_name:
+    value: box-close-v2
+training:
+    value:
+        checkpoint_every: 100
+        debug: false
+        device: cuda:0
+        grad_norm_clip: 0.5
+        gradient_accumulate_every: 8
+        lr_scheduler: cosine
+        lr_warmup_steps: 10
+        max_train_steps: null
+        max_val_steps: null
+        num_epochs: 20001
+        resume: false
+        rollout_every: 100
+        sample_every: 100
+        sample_max_batch: 128
+        seed: 43
+        tqdm_interval_sec: 1
+        val_every: 1
+val_dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: true

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/files/output.log

@@ -0,0 +1,3 @@
+Eval MetaworldLowdimRunner 1/6:   0%|          | 0/30 [00:00<?, ?it/s]/work/u1131674/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)
+                                                                                                      

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/files/wandb-metadata.json

@@ -0,0 +1,107 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.28",
+  "python": "3.9.25",
+  "startedAt": "2026-01-27T03:53:08.073313Z",
+  "args": [
+    "--config-path",
+    "config/data_efficiency",
+    "--config-name",
+    "llmbc_box-close-v2_o4-mini_split0.05.yaml",
+    "training.seed=43"
+  ],
+  "program": "/work/u1131674/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "206b9c57fa687707dfa1ee0ebb33b507135a59c6"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2",
+  "host": "hgpn06",
+  "username": "u1131674",
+  "executable": "/home/u1131674/.conda/envs/llm-bc/bin/python3",
+  "codePathLocal": "train.py",
+  "cpu_count": 112,
+  "cpu_count_logical": 112,
+  "gpu": "NVIDIA H100 80GB HBM3",
+  "gpu_count": 1,
+  "disk": {
+    "/": {
+      "total": "1918024196096",
+      "used": "333231804416"
+    }
+  },
+  "memory": {
+    "total": "2163623079936"
+  },
+  "cpu": {
+    "count": 112,
+    "countLogical": 112
+  },
+  "gpu_nvidia": [
+    {
+      "name": "NVIDIA H100 80GB HBM3",
+      "memoryTotal": "85520809984",
+      "cudaCores": 16896,
+      "architecture": "Hopper"
+    }
+  ],
+  "slurm": {
+    "cluster_name": "hpc",
+    "conf": "/etc/slurm/slurm.conf",
+    "cpu_bind": "quiet,mask_cpu:0x0000000000000000000000F0000F",
+    "cpu_bind_list": "0x0000000000000000000000F0000F",
+    "cpu_bind_type": "mask_cpu:",
+    "cpu_bind_verbose": "quiet",
+    "cpus_on_node": "8",
+    "cpus_per_task": "8",
+    "distribution": "cyclic,pack",
+    "gpus_on_node": "1",
+    "gpus_per_node": "1",
+    "gtids": "0",
+    "job_account": "mst114558",
+    "job_cpus_per_node": "8",
+    "job_end_time": "1769658768",
+    "job_gid": "106773",
+    "job_group": "MST114558",
+    "job_id": "104243",
+    "job_name": "python3",
+    "job_nodelist": "hgpn06",
+    "job_num_nodes": "1",
+    "job_partition": "normal",
+    "job_qos": "normal",
+    "job_start_time": "1769485968",
+    "job_uid": "41408",
+    "job_user": "u1131674",
+    "jobid": "104243",
+    "launch_node_ipaddr": "172.21.101.1",
+    "localid": "0",
+    "mem_per_node": "204800",
+    "nnodes": "1",
+    "nodeid": "0",
+    "nodelist": "hgpn06",
+    "nprocs": "1",
+    "ntasks": "1",
+    "prio_process": "0",
+    "procid": "0",
+    "srun_comm_host": "172.21.101.1",
+    "srun_comm_port": "44181",
+    "step_gpus": "3",
+    "step_id": "0",
+    "step_launcher_port": "44181",
+    "step_nodelist": "hgpn06",
+    "step_num_nodes": "1",
+    "step_num_tasks": "1",
+    "step_tasks_per_node": "1",
+    "stepid": "0",
+    "submit_dir": "/work/u1131674/LLM-BC",
+    "submit_host": "cbi-lgn01",
+    "task_pid": "2419965",
+    "tasks_per_node": "1",
+    "topology_addr": "ibsw1.hgpn06",
+    "topology_addr_pattern": "switch.node",
+    "tres_per_task": "cpu:8",
+    "umask": "0022"
+  },
+  "cudaVersion": "12.4"
+}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/files/wandb-summary.json

@@ -0,0 +1 @@
+{"_timestamp":1.769524654311829e+09,"global_step":260012,"test/success_rate":0.54,"grad_norm":4.1202642023563385e-05,"lr":2.337295357257574e-11,"_runtime":38668.49954648,"train_loss_llm":9.465217590332031e-05,"test/cumulative_reward":12.559526529061893,"train/success_rate":0.8,"val_loss":2.897437116189394e-05,"train_action_mse_error":1.0403916661516632e-07,"train/cumulative_reward":14.612570521362438,"train/mean_score":6.996024131036938,"test/mean_score":5.711785312482341,"_wandb":{"runtime":38668},"epoch":20000,"train_loss":1.5365032077432681e-07,"_step":260012,"train_loss_bc":9.23023861787442e-08}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug-core.log

@@ -0,0 +1,14 @@
+{"time":"2026-01-27T11:53:07.47897058+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpy8hm3v8r/port-2419965.txt","pid":2419965,"debug":false,"disable-analytics":false}
+{"time":"2026-01-27T11:53:07.478992022+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-27T11:53:07.479516556+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":2419965}
+{"time":"2026-01-27T11:53:07.479510234+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":40351,"Zone":""}}
+{"time":"2026-01-27T11:53:07.672421694+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T11:53:08.073873525+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"raachjxu","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T11:53:08.190317831+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"raachjxu","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T22:37:36.572632384+08:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T22:37:36.572847244+08:00","level":"INFO","msg":"server is shutting down"}
+{"time":"2026-01-27T22:37:36.572836178+08:00","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T22:37:36.572940064+08:00","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T22:37:41.125798856+08:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T22:37:41.125815466+08:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:60334"}
+{"time":"2026-01-27T22:37:41.125826131+08:00","level":"INFO","msg":"server is closed"}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug-internal.log

@@ -0,0 +1,16 @@
+{"time":"2026-01-27T11:53:08.074868883+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-27T11:53:08.074878267+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug-core.log"}
+{"time":"2026-01-27T11:53:08.190280513+08:00","level":"INFO","msg":"created new stream","id":"raachjxu"}
+{"time":"2026-01-27T11:53:08.190312975+08:00","level":"INFO","msg":"stream: started","id":"raachjxu"}
+{"time":"2026-01-27T11:53:08.190338185+08:00","level":"INFO","msg":"sender: started","stream_id":"raachjxu"}
+{"time":"2026-01-27T11:53:08.190329168+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T11:53:08.1903328+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T11:53:08.732831596+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-01-27T22:37:36.57283309+08:00","level":"INFO","msg":"stream: closing","id":"raachjxu"}
+{"time":"2026-01-27T22:37:36.572875878+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-01-27T22:37:36.573489042+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-01-27T22:37:40.850119587+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-01-27T22:37:41.125467393+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T22:37:41.125498907+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"raachjxu"}}
+{"time":"2026-01-27T22:37:41.125540962+08:00","level":"INFO","msg":"sender: closed","stream_id":"raachjxu"}
+{"time":"2026-01-27T22:37:41.125562117+08:00","level":"INFO","msg":"stream: closed","id":"raachjxu"}

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug.log

@@ -0,0 +1,28 @@
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Configure stats pid to 2419965
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-27 11:53:08,070 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/work/u1131674/LLM-BC/train.py', 'program': '/work/u1131674/LLM-BC/./train.py'}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug.log
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/logs/debug-internal.log
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():619] calling init triggers
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llmbc_lowdim', '_target_': 'llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace', 'obs_dim': 9, 'action_dim': 4, 'task_name': 'box-close-v2', 'exp_name': 'default', 'model_name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy', 'model': {'_target_': 'llmbc.model.policy.policy_mlp.PolicyMLP', 'input_size': 9, 'hidden_size': [256, 256], 'output_size': 4, 'activation': 'relu', 'n_obs_steps': 1, 'n_action_steps': 1}, 'obs_dim': 9, 'action_dim': 4, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'box-close-v2', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_bc_weight': 1.0, 'loss_llm_weight': 0.001, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'normalize_llm_loss': True}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.01, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 43, 'debug': False, 'resume': False, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 20001, 'gradient_accumulate_every': 8, 'grad_norm_clip': 0.5, 'rollout_every': 100, 'checkpoint_every': 100, 'val_every': 1, 'sample_every': 100, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'box-close-v2-training-split-ablation', 'resume': True, 'mode': 'online', 'name': '2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2', 'tags': ['train_llmbc_lowdim', 'box-close-v2', '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/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2', 'wandb_name_base': '2026.01.27-11.52.50_train_llmbc_lowdim_box-close-v2'}, 'task': {'name': 'box-close-v2', 'obs_dim': 9, 'action_dim': 4, 'env_runner': {'_target_': 'llmbc.env_runner.metaworld_lowdim_runner.MetaworldLowdimRunner', 'env_name': 'llf-metaworld-box-close-v2', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 30, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.9}, 'dataset': {'_target_': 'llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset', 'data_path': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'data_path2': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': True}, 'instructor': {'_target_': 'llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.52.50_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():669] starting backend
+2026-01-27 11:53:08,071 INFO    MainThread:2419965 [wandb_init.py:init():673] sending inform_init request
+2026-01-27 11:53:08,072 INFO    MainThread:2419965 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-27 11:53:08,073 INFO    MainThread:2419965 [wandb_init.py:init():686] backend started and connected
+2026-01-27 11:53:08,082 INFO    MainThread:2419965 [wandb_init.py:init():781] updated telemetry
+2026-01-27 11:53:08,142 INFO    MainThread:2419965 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-27 11:53:08,717 INFO    MainThread:2419965 [wandb_init.py:init():859] run resumed
+2026-01-27 11:53:08,728 INFO    MainThread:2419965 [wandb_init.py:init():867] starting run threads in backend
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-27 11:53:09,276 INFO    MainThread:2419965 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-27 11:53:09,294 INFO    MainThread:2419965 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-27 11:53:09,295 INFO    MainThread:2419965 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2'}
+2026-01-27 22:37:36,572 WARNING MsgRouterThr:2419965 [router.py:message_loop():75] message_loop has been closed

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115308-raachjxu/run-raachjxu.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3781f811f713b9c3f856e19b8a57bcc0d88ac14e4012b98457d6016ab9513a48
+size 218982786

2026.01.27/11.52.50_train_llmbc_lowdim_box-close-v2/wandb/wandb-resume.json

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

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/.hydra/config.yaml

@@ -0,0 +1,163 @@
+name: train_llmbc_lowdim
+_target_: llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace
+obs_dim: ${task.obs_dim}
+action_dim: ${task.action_dim}
+task_name: ${task.name}
+exp_name: default
+model_name: ${llm.name}
+horizon: 1
+n_obs_steps: 1
+n_action_steps: 1
+n_latency_steps: 0
+past_action_visible: false
+llm_orig_expert_feedback: true
+llm_do_sample: false
+policy:
+  _target_: llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy
+  model:
+    _target_: llmbc.model.policy.policy_mlp.PolicyMLP
+    input_size: ${eval:'${n_obs_steps}*${obs_dim}'}
+    hidden_size:
+    - 256
+    - 256
+    output_size: ${eval:'${n_action_steps}*${action_dim}'}
+    activation: relu
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+  obs_dim: ${obs_dim}
+  action_dim: ${action_dim}
+  llm_discriminator:
+    _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+    task_id: ${task_name}
+    llm_translator:
+      _target_: llmbc.translator.llm_translator.LLMTranslator
+      cfg: ${llm}
+      obs_dim: ${task.obs_dim}
+      action_dim: ${task.action_dim}
+      horizon: ${horizon}
+      n_obs_steps: ${n_obs_steps}
+      n_action_steps: ${n_action_steps}
+  loss_bc_weight: 1.0
+  loss_llm_weight: 0.001
+  horizon: ${horizon}
+  n_obs_steps: ${n_obs_steps}
+  n_action_steps: ${n_action_steps}
+  normalize_llm_loss: true
+dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.AdamW
+  lr: 0.01
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda:0
+  seed: 42
+  debug: false
+  resume: false
+  lr_scheduler: cosine
+  lr_warmup_steps: 10
+  num_epochs: 20001
+  gradient_accumulate_every: 8
+  grad_norm_clip: 0.5
+  rollout_every: 100
+  checkpoint_every: 100
+  val_every: 1
+  sample_every: 100
+  sample_max_batch: 128
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: ${task.name}-training-split-ablation
+  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: box-close-v2
+  obs_dim: 9
+  action_dim: 4
+  env_runner:
+    _target_: llmbc.env_runner.metaworld_lowdim_runner.MetaworldLowdimRunner
+    env_name: llf-metaworld-box-close-v2
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 30
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    - fp
+    visual: false
+    discount: 0.9
+  dataset:
+    _target_: llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset
+    data_path: datasets/box-close-v2-o4-mini-split0.05.pt
+    data_path2: datasets/box-close-v2-o4-mini-split0.05.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: true
+  instructor:
+    _target_: llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor
+llm:
+  name: HuggingFaceTB/SmolLM2-135M-Instruct
+  model_name: SmolLM2-135M-Instruct
+  config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+  causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+  use_quantization: false
+  use_joint_mlp_projector: true
+  llm_mode: ete-finetuned
+  finetune_mode: orig
+  checkpoint: data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000
+  max_length: 100
+  lora_config:
+    r: 32
+    lora_alpha: 64
+    lora_dropout: 0.05
+    bias: none
+    task_type: CAUSAL_LM
+  prompter:
+    _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+    use_joint_mlp_projector: true
+  hydra:
+    job:
+      override_dirname: ${model_name}
+    run:
+      dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${model_name}

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+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=42
+  job:
+    name: train
+    chdir: null
+    override_dirname: training.seed=42
+    id: ???
+    num: ???
+    config_name: llmbc_box-close-v2_o4-mini_split0.05.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /work/u1131674/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /work/u1131674/LLM-BC/config/data_efficiency
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2
+    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

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- training.seed=42

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=10200-test_success_rate=0.780.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8c36a870df18b492f96b325312fc3238fce7982d9fc259e8970d1179c41132c
+size 864584

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=11400-test_success_rate=0.800.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bd584f31532b75397ab25728c564c692c62f02ca3fa992b538b3ca13a1a0090
+size 864584

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=12200-test_success_rate=0.800.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b7e34cb6bba63eff86f3cf8878f4a0932b7f06f56f4c8d3cde6182a3c92d528
+size 864584

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=12300-test_success_rate=0.780.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3392da6ccd4ac2d41f68d8f6377794f15554518a67fdb6a0d3a8052508e8b4a2
+size 864584

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/checkpoints/epoch=7400-test_success_rate=0.780.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0e75ab803b9efb1e8638d92610ea02fba48a06fd5506a581fa1879d31c77210
+size 864584

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/checkpoints/latest.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ab07a2c125d3a3be246065ef60aaf6cfafa80726d8e1b3fe770559db36dbc0e
+size 864584

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/logs.json.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25aa1e99560ada5a21c6532ce1106f103ee143edcc7b42ef3f34dd1d9f1ea0bb
+size 55577527

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/train.log

@@ -0,0 +1,8 @@
+[2026-01-27 11:59:33,618][numexpr.utils][INFO] - Note: detected 112 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2026-01-27 11:59:33,618][numexpr.utils][INFO] - Note: NumExpr detected 112 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2026-01-27 11:59:33,618][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2026-01-27 11:59:38,326][datasets][INFO] - PyTorch version 2.2.2 available.
+[2026-01-27 11:59:38,327][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2026-01-27 11:59:38,327][datasets][INFO] - JAX version 0.4.30 available.
+[2026-01-27 11:59:45,313][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
+[2026-01-27 11:59:45,322][absl][INFO] - MuJoCo library version is: 2.3.7

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/debug-internal.log

@@ -0,0 +1,16 @@
+{"time":"2026-01-27T11:59:46.233224045+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-27T11:59:46.233232608+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/logs/debug-core.log"}
+{"time":"2026-01-27T11:59:46.347396478+08:00","level":"INFO","msg":"created new stream","id":"ccfd9xxl"}
+{"time":"2026-01-27T11:59:46.347428899+08:00","level":"INFO","msg":"stream: started","id":"ccfd9xxl"}
+{"time":"2026-01-27T11:59:46.347452905+08:00","level":"INFO","msg":"sender: started","stream_id":"ccfd9xxl"}
+{"time":"2026-01-27T11:59:46.347448548+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T11:59:46.347443454+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T11:59:47.036042163+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-01-27T22:47:47.651604587+08:00","level":"INFO","msg":"stream: closing","id":"ccfd9xxl"}
+{"time":"2026-01-27T22:47:47.651640552+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-01-27T22:47:47.65220418+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-01-27T22:47:48.903771416+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-01-27T22:47:49.193540342+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T22:47:49.19356658+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T22:47:49.193600819+08:00","level":"INFO","msg":"sender: closed","stream_id":"ccfd9xxl"}
+{"time":"2026-01-27T22:47:49.19361559+08:00","level":"INFO","msg":"stream: closed","id":"ccfd9xxl"}

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/debug.log

@@ -0,0 +1,27 @@
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Configure stats pid to 888299
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/work/u1131674/LLM-BC/train.py', 'program': '/work/u1131674/LLM-BC/./train.py'}
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/logs/debug.log
+2026-01-27 11:59:46,229 INFO    MainThread:888299 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/logs/debug-internal.log
+2026-01-27 11:59:46,230 INFO    MainThread:888299 [wandb_init.py:init():619] calling init triggers
+2026-01-27 11:59:46,230 INFO    MainThread:888299 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llmbc_lowdim', '_target_': 'llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace', 'obs_dim': 9, 'action_dim': 4, 'task_name': 'box-close-v2', 'exp_name': 'default', 'model_name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'n_latency_steps': 0, 'past_action_visible': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy', 'model': {'_target_': 'llmbc.model.policy.policy_mlp.PolicyMLP', 'input_size': 9, 'hidden_size': [256, 256], 'output_size': 4, 'activation': 'relu', 'n_obs_steps': 1, 'n_action_steps': 1}, 'obs_dim': 9, 'action_dim': 4, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'box-close-v2', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.59.31_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_bc_weight': 1.0, 'loss_llm_weight': 0.001, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'normalize_llm_loss': True}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.01, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 42, 'debug': False, 'resume': False, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 20001, 'gradient_accumulate_every': 8, 'grad_norm_clip': 0.5, 'rollout_every': 100, 'checkpoint_every': 100, 'val_every': 1, 'sample_every': 100, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'box-close-v2-training-split-ablation', 'resume': True, 'mode': 'online', 'name': '2026.01.27-11.59.31_train_llmbc_lowdim_box-close-v2', 'tags': ['train_llmbc_lowdim', 'box-close-v2', '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/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2', 'wandb_name_base': '2026.01.27-11.59.31_train_llmbc_lowdim_box-close-v2'}, 'task': {'name': 'box-close-v2', 'obs_dim': 9, 'action_dim': 4, 'env_runner': {'_target_': 'llmbc.env_runner.metaworld_lowdim_runner.MetaworldLowdimRunner', 'env_name': 'llf-metaworld-box-close-v2', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 30, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.9}, 'dataset': {'_target_': 'llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset', 'data_path': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'data_path2': 'datasets/box-close-v2-o4-mini-split0.05.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.1, 'dummy_normalizer': True}, 'instructor': {'_target_': 'llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.01.27/11.59.31_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-27 11:59:46,230 INFO    MainThread:888299 [wandb_init.py:init():669] starting backend
+2026-01-27 11:59:46,230 INFO    MainThread:888299 [wandb_init.py:init():673] sending inform_init request
+2026-01-27 11:59:46,231 INFO    MainThread:888299 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-27 11:59:46,231 INFO    MainThread:888299 [wandb_init.py:init():686] backend started and connected
+2026-01-27 11:59:46,240 INFO    MainThread:888299 [wandb_init.py:init():781] updated telemetry
+2026-01-27 11:59:46,295 INFO    MainThread:888299 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-27 11:59:47,031 INFO    MainThread:888299 [wandb_init.py:init():867] starting run threads in backend
+2026-01-27 11:59:47,512 INFO    MainThread:888299 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-27 11:59:47,512 INFO    MainThread:888299 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-27 11:59:47,512 INFO    MainThread:888299 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-27 11:59:47,512 INFO    MainThread:888299 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-27 11:59:47,514 INFO    MainThread:888299 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-27 11:59:47,514 INFO    MainThread:888299 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2'}
+2026-01-27 22:47:47,651 WARNING MsgRouterThr:888299 [router.py:message_loop():75] message_loop has been closed

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/files/config.yaml

@@ -0,0 +1,271 @@
+_target_:
+    value: llmbc.workspace.train_llmbc_lowdim_workspace.TrainLLMBCLowdimWorkspace
+_wandb:
+    value:
+        cli_version: 0.18.6
+        m: []
+        python_version: 3.9.25
+        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": 3.9.25
+            "5": 0.18.6
+            "6": 4.47.1
+            "8":
+                - 5
+            "12": 0.18.6
+            "13": linux-x86_64
+action_dim:
+    value: 4
+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
+dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: true
+exp_name:
+    value: default
+horizon:
+    value: 1
+llm:
+    value:
+        causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+        checkpoint: data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000
+        config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+        finetune_mode: orig
+        hydra:
+            job:
+                override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+            run:
+                dir: data/outputs/2026.01.27/11.59.31_HuggingFaceTB/SmolLM2-135M-Instruct
+        llm_mode: ete-finetuned
+        lora_config:
+            bias: none
+            lora_alpha: 64
+            lora_dropout: 0.05
+            r: 32
+            task_type: CAUSAL_LM
+        max_length: 100
+        model_name: SmolLM2-135M-Instruct
+        name: HuggingFaceTB/SmolLM2-135M-Instruct
+        prompter:
+            _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+            use_joint_mlp_projector: true
+        use_joint_mlp_projector: true
+        use_quantization: false
+llm_do_sample:
+    value: false
+llm_orig_expert_feedback:
+    value: true
+logging:
+    value:
+        group: null
+        id: null
+        mode: online
+        name: 2026.01.27-11.59.31_train_llmbc_lowdim_box-close-v2
+        project: box-close-v2-training-split-ablation
+        resume: true
+        tags:
+            - train_llmbc_lowdim
+            - box-close-v2
+            - default
+model_name:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+multi_run:
+    value:
+        run_dir: data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2
+        wandb_name_base: 2026.01.27-11.59.31_train_llmbc_lowdim_box-close-v2
+n_action_steps:
+    value: 1
+n_latency_steps:
+    value: 0
+n_obs_steps:
+    value: 1
+name:
+    value: train_llmbc_lowdim
+obs_dim:
+    value: 9
+optimizer:
+    value:
+        _target_: torch.optim.AdamW
+        betas:
+            - 0.95
+            - 0.999
+        eps: 1e-08
+        lr: 0.01
+        weight_decay: 1e-06
+output_dir:
+    value: /work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2
+past_action_visible:
+    value: false
+policy:
+    value:
+        _target_: llmbc.policy.llmbc_lowdim_policy.LLMBCLowdimPolicy
+        action_dim: 4
+        horizon: 1
+        llm_discriminator:
+            _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+            llm_translator:
+                _target_: llmbc.translator.llm_translator.LLMTranslator
+                action_dim: 4
+                cfg:
+                    causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+                    checkpoint: data/outputs/2026.01.26/10.26.52_train_llm_lowdim_box-close-v2/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-box-close-v2/checkpoint-6000
+                    config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+                    finetune_mode: orig
+                    hydra:
+                        job:
+                            override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+                        run:
+                            dir: data/outputs/2026.01.27/11.59.31_HuggingFaceTB/SmolLM2-135M-Instruct
+                    llm_mode: ete-finetuned
+                    lora_config:
+                        bias: none
+                        lora_alpha: 64
+                        lora_dropout: 0.05
+                        r: 32
+                        task_type: CAUSAL_LM
+                    max_length: 100
+                    model_name: SmolLM2-135M-Instruct
+                    name: HuggingFaceTB/SmolLM2-135M-Instruct
+                    prompter:
+                        _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+                        use_joint_mlp_projector: true
+                    use_joint_mlp_projector: true
+                    use_quantization: false
+                horizon: 1
+                n_action_steps: 1
+                n_obs_steps: 1
+                obs_dim: 9
+            task_id: box-close-v2
+        loss_bc_weight: 1
+        loss_llm_weight: 0.001
+        model:
+            _target_: llmbc.model.policy.policy_mlp.PolicyMLP
+            activation: relu
+            hidden_size:
+                - 256
+                - 256
+            input_size: 9
+            n_action_steps: 1
+            n_obs_steps: 1
+            output_size: 4
+        n_action_steps: 1
+        n_obs_steps: 1
+        normalize_llm_loss: true
+        obs_dim: 9
+task:
+    value:
+        action_dim: 4
+        dataset:
+            _target_: llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset
+            data_path: datasets/box-close-v2-o4-mini-split0.05.pt
+            data_path2: datasets/box-close-v2-o4-mini-split0.05.pt
+            dummy_normalizer: true
+            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.metaworld_lowdim_runner.MetaworldLowdimRunner
+            discount: 0.9
+            env_name: llf-metaworld-box-close-v2
+            feedback_type:
+                - hp
+                - hn
+                - fp
+            instruction_type: b
+            max_steps: 30
+            n_action_steps: 1
+            n_envs: 10
+            n_obs_steps: 1
+            n_test: 50
+            n_train: 10
+            visual: false
+        instructor:
+            _target_: llmbc.translator.instructor.metaworld_instructor.box_close_v2_instructor.BoxCloseV2Instructor
+        name: box-close-v2
+        obs_dim: 9
+task_name:
+    value: box-close-v2
+training:
+    value:
+        checkpoint_every: 100
+        debug: false
+        device: cuda:0
+        grad_norm_clip: 0.5
+        gradient_accumulate_every: 8
+        lr_scheduler: cosine
+        lr_warmup_steps: 10
+        max_train_steps: null
+        max_val_steps: null
+        num_epochs: 20001
+        resume: false
+        rollout_every: 100
+        sample_every: 100
+        sample_max_batch: 128
+        seed: 42
+        tqdm_interval_sec: 1
+        val_every: 1
+val_dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: true

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/files/output.log

@@ -0,0 +1,3 @@
+Eval MetaworldLowdimRunner 1/6:   0%|          | 0/30 [00:00<?, ?it/s]/work/u1131674/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)
+                                                                                                       

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/files/wandb-metadata.json

@@ -0,0 +1,107 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.28",
+  "python": "3.9.25",
+  "startedAt": "2026-01-27T03:59:46.231929Z",
+  "args": [
+    "--config-path",
+    "config/data_efficiency",
+    "--config-name",
+    "llmbc_box-close-v2_o4-mini_split0.05.yaml",
+    "training.seed=42"
+  ],
+  "program": "/work/u1131674/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "206b9c57fa687707dfa1ee0ebb33b507135a59c6"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2",
+  "host": "hgpn05",
+  "username": "u1131674",
+  "executable": "/home/u1131674/.conda/envs/llm-bc/bin/python3",
+  "codePathLocal": "train.py",
+  "cpu_count": 112,
+  "cpu_count_logical": 112,
+  "gpu": "NVIDIA H100 80GB HBM3",
+  "gpu_count": 1,
+  "disk": {
+    "/": {
+      "total": "1918024196096",
+      "used": "437314154496"
+    }
+  },
+  "memory": {
+    "total": "2163617771520"
+  },
+  "cpu": {
+    "count": 112,
+    "countLogical": 112
+  },
+  "gpu_nvidia": [
+    {
+      "name": "NVIDIA H100 80GB HBM3",
+      "memoryTotal": "85520809984",
+      "cudaCores": 16896,
+      "architecture": "Hopper"
+    }
+  ],
+  "slurm": {
+    "cluster_name": "hpc",
+    "conf": "/etc/slurm/slurm.conf",
+    "cpu_bind": "quiet,mask_cpu:0x000000000FE04000000000000000",
+    "cpu_bind_list": "0x000000000FE04000000000000000",
+    "cpu_bind_type": "mask_cpu:",
+    "cpu_bind_verbose": "quiet",
+    "cpus_on_node": "8",
+    "cpus_per_task": "8",
+    "distribution": "cyclic,pack",
+    "gpus_on_node": "1",
+    "gpus_per_node": "1",
+    "gtids": "0",
+    "job_account": "mst114558",
+    "job_cpus_per_node": "8",
+    "job_end_time": "1769659170",
+    "job_gid": "106773",
+    "job_group": "MST114558",
+    "job_id": "104250",
+    "job_name": "python3",
+    "job_nodelist": "hgpn05",
+    "job_num_nodes": "1",
+    "job_partition": "normal",
+    "job_qos": "normal",
+    "job_start_time": "1769486370",
+    "job_uid": "41408",
+    "job_user": "u1131674",
+    "jobid": "104250",
+    "launch_node_ipaddr": "172.21.101.1",
+    "localid": "0",
+    "mem_per_node": "204800",
+    "nnodes": "1",
+    "nodeid": "0",
+    "nodelist": "hgpn05",
+    "nprocs": "1",
+    "ntasks": "1",
+    "prio_process": "0",
+    "procid": "0",
+    "srun_comm_host": "172.21.101.1",
+    "srun_comm_port": "43119",
+    "step_gpus": "5",
+    "step_id": "0",
+    "step_launcher_port": "43119",
+    "step_nodelist": "hgpn05",
+    "step_num_nodes": "1",
+    "step_num_tasks": "1",
+    "step_tasks_per_node": "1",
+    "stepid": "0",
+    "submit_dir": "/work/u1131674/LLM-BC",
+    "submit_host": "cbi-lgn01",
+    "task_pid": "888299",
+    "tasks_per_node": "1",
+    "topology_addr": "ibsw1.hgpn05",
+    "topology_addr_pattern": "switch.node",
+    "tres_per_task": "cpu:8",
+    "umask": "0022"
+  },
+  "cudaVersion": "12.4"
+}

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/files/wandb-summary.json

@@ -0,0 +1 @@
+{"train/cumulative_reward":11.42187131867422,"_runtime":38881.419697085,"test/success_rate":0.6,"_wandb":{"runtime":38881},"train_action_mse_error":2.106957985859026e-08,"val_loss":0.00045819077058695257,"grad_norm":1.85015014722012e-05,"global_step":260012,"epoch":20000,"train_loss_llm":0,"test/mean_score":5.574101696679537,"test/cumulative_reward":12.715780319485676,"train_loss_bc":8.472071755249999e-08,"_step":260012,"_timestamp":1.7695252651366467e+09,"lr":2.337295357257574e-11,"train/success_rate":0.5,"train/mean_score":4.583878996955714,"train_loss":8.182333484267718e-08}

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/logs/debug-core.log

@@ -0,0 +1,14 @@
+{"time":"2026-01-27T11:59:45.620925943+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpipd6umd5/port-888299.txt","pid":888299,"debug":false,"disable-analytics":false}
+{"time":"2026-01-27T11:59:45.620947818+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-27T11:59:45.621295865+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":888299}
+{"time":"2026-01-27T11:59:45.621289332+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":41057,"Zone":""}}
+{"time":"2026-01-27T11:59:45.815580254+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T11:59:46.232446283+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"ccfd9xxl","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T11:59:46.347434312+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"ccfd9xxl","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T22:47:47.651402799+08:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T22:47:47.65160313+08:00","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T22:47:47.651669651+08:00","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T22:47:47.651612942+08:00","level":"INFO","msg":"server is shutting down"}
+{"time":"2026-01-27T22:47:49.193798825+08:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T22:47:49.193814452+08:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:33124"}
+{"time":"2026-01-27T22:47:49.193826036+08:00","level":"INFO","msg":"server is closed"}

2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/logs/debug-internal.log

@@ -0,0 +1,16 @@
+{"time":"2026-01-27T11:59:46.233224045+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-27T11:59:46.233232608+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.27/11.59.31_train_llmbc_lowdim_box-close-v2/wandb/run-20260127_115946-ccfd9xxl/logs/debug-core.log"}
+{"time":"2026-01-27T11:59:46.347396478+08:00","level":"INFO","msg":"created new stream","id":"ccfd9xxl"}
+{"time":"2026-01-27T11:59:46.347428899+08:00","level":"INFO","msg":"stream: started","id":"ccfd9xxl"}
+{"time":"2026-01-27T11:59:46.347452905+08:00","level":"INFO","msg":"sender: started","stream_id":"ccfd9xxl"}
+{"time":"2026-01-27T11:59:46.347448548+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T11:59:46.347443454+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T11:59:47.036042163+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-01-27T22:47:47.651604587+08:00","level":"INFO","msg":"stream: closing","id":"ccfd9xxl"}
+{"time":"2026-01-27T22:47:47.651640552+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-01-27T22:47:47.65220418+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-01-27T22:47:48.903771416+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-01-27T22:47:49.193540342+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T22:47:49.19356658+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"ccfd9xxl"}}
+{"time":"2026-01-27T22:47:49.193600819+08:00","level":"INFO","msg":"sender: closed","stream_id":"ccfd9xxl"}
+{"time":"2026-01-27T22:47:49.19361559+08:00","level":"INFO","msg":"stream: closed","id":"ccfd9xxl"}