Upload folder using huggingface_hub

.gitattributes

@@ -47,3 +47,12 @@ saved_model/**/* 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
+2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/run-d9kt1a6y.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.28/00.21.32_train_mlp_projector_box-close-v2/wandb/run-20260128_002145-uvg24ws0/run-uvg24ws0.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.28/11.01.00_train_llm_lowdim_box-close-v2/wandb/run-20260128_110109-g0yynelh/run-g0yynelh.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.28/11.02.01_train_llm_lowdim_box-close-v2/wandb/run-20260128_110208-3y5kjuul/run-3y5kjuul.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.28/11.51.52_train_llm_lowdim_box-close-v2/wandb/run-20260128_115158-aqxmr1ik/run-aqxmr1ik.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.28/12.27.30_train_llm_diffusion_unet_lowdim_box-close-v2/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2026.01.28/12.27.30_train_llm_diffusion_unet_lowdim_box-close-v2/wandb/run-20260128_122749-52m9owgb/run-52m9owgb.wandb filter=lfs diff=lfs merge=lfs -text
+2026.01.28/12.35.32_train_llm_diffusion_unet_lowdim_box-close-v2/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2026.01.28/12.35.32_train_llm_diffusion_unet_lowdim_box-close-v2/wandb/run-20260128_123551-yt3zq46y/run-yt3zq46y.wandb filter=lfs diff=lfs merge=lfs -text

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/.hydra/config.yaml

@@ -0,0 +1,139 @@
+name: train_mlp_projector
+_target_: llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace
+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_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}
+dataloader:
+  batch_size: 128
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 128
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.Adam
+  lr: 0.0001
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda
+  seed: 42
+  debug: false
+  resume: true
+  lr_scheduler: cosine
+  lr_warmup_steps: 10
+  num_epochs: 100
+  gradient_accumulate_every: 2
+  grad_norm_clip: 1
+  enable_normalizer: true
+  checkpoint_every: 1
+  val_every: 1
+  sample_every: 1
+  sample_max_batch: 128
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: llm_module_training
+  resume: true
+  mode: online
+  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+  tags:
+  - ${name}
+  - ${task_name}
+  - ${exp_name}
+  id: null
+  group: null
+checkpoint:
+  topk:
+    monitor_key: val_loss
+    mode: min
+    k: 5
+    format_str: epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt
+  save_last_ckpt: true
+  save_last_snapshot: false
+multi_run:
+  run_dir: data/outputs/${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-general-o4-mini.pt
+    data_path2: datasets/box-close-v2-o4-mini.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: 0.05
+    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: pretrained
+  finetune_mode: orig
+  checkpoint: ''
+  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}
+

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/.hydra/hydra.yaml

@@ -0,0 +1,154 @@
+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: []
+  job:
+    name: train
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: llmbc_mlp_projector_box-close-v2_o4-mini.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/lang_source
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_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

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/logs.json.txt

@@ -0,0 +1,67 @@
+{"train_loss": 3.8194987773895264, "grad_norm": 0.5754198431968689, "global_step": 0, "epoch": 0, "lr": 1e-05}
+{"train_loss": 3.7512643337249756, "grad_norm": 0.7266139388084412, "global_step": 1, "epoch": 0, "lr": 1e-05}
+{"train_loss": 3.815833330154419, "grad_norm": 0.0843488797545433, "global_step": 0, "epoch": 0, "lr": 1e-05}
+{"train_loss": 3.7482757568359375, "grad_norm": 0.07725829631090164, "global_step": 1, "epoch": 0, "lr": 1e-05}
+{"train_loss": 3.8328254222869873, "grad_norm": 0.19070754945278168, "global_step": 2, "epoch": 0, "lr": 2e-05}
+{"train_loss": 3.8107516765594482, "grad_norm": 0.09841667860746384, "global_step": 3, "epoch": 0, "lr": 2e-05}
+{"train_loss": 3.789482355117798, "grad_norm": 0.23507383465766907, "global_step": 4, "epoch": 0, "lr": 3e-05}
+{"train_loss": 3.7551498413085938, "grad_norm": 0.08915789425373077, "global_step": 5, "epoch": 0, "lr": 3e-05}
+{"train_loss": 3.764460563659668, "grad_norm": 0.2220054566860199, "global_step": 6, "epoch": 0, "lr": 4e-05}
+{"train_loss": 3.80804705619812, "grad_norm": 0.11431895941495895, "global_step": 7, "epoch": 0, "lr": 4e-05}
+{"train_loss": 3.8422603607177734, "grad_norm": 0.22483113408088684, "global_step": 8, "epoch": 0, "lr": 5e-05}
+{"train_loss": 3.7713098526000977, "grad_norm": 0.12273082882165909, "global_step": 9, "epoch": 0, "lr": 5e-05}
+{"train_loss": 3.771944284439087, "grad_norm": 0.19527220726013184, "global_step": 10, "epoch": 0, "lr": 6e-05}
+{"train_loss": 3.746992826461792, "grad_norm": 0.15018615126609802, "global_step": 11, "epoch": 0, "lr": 6e-05}
+{"train_loss": 3.7489941120147705, "grad_norm": 0.21353769302368164, "global_step": 12, "epoch": 0, "lr": 7e-05}
+{"train_loss": 3.843421220779419, "grad_norm": 0.10667309910058975, "global_step": 13, "epoch": 0, "lr": 7e-05}
+{"train_loss": 3.782552480697632, "grad_norm": 0.20214207470417023, "global_step": 14, "epoch": 0, "lr": 8e-05}
+{"train_loss": 3.7685015201568604, "grad_norm": 0.1669682413339615, "global_step": 15, "epoch": 0, "lr": 8e-05}
+{"train_loss": 3.815293073654175, "grad_norm": 0.2628539800643921, "global_step": 16, "epoch": 0, "lr": 9e-05}
+{"train_loss": 3.7662413120269775, "grad_norm": 0.15504860877990723, "global_step": 17, "epoch": 0, "lr": 9e-05}
+{"train_loss": 3.795480251312256, "grad_norm": 0.3150642514228821, "global_step": 18, "epoch": 0, "lr": 0.0001}
+{"train_loss": 3.8038082122802734, "grad_norm": 0.17351281642913818, "global_step": 19, "epoch": 0, "lr": 0.0001}
+{"train_loss": 3.799712657928467, "grad_norm": 0.2751327157020569, "global_step": 20, "epoch": 0, "lr": 9.99999996906071e-05}
+{"train_loss": 3.7262821197509766, "grad_norm": 0.21049296855926514, "global_step": 21, "epoch": 0, "lr": 9.99999996906071e-05}
+{"train_loss": 3.729520320892334, "grad_norm": 0.2575721740722656, "global_step": 22, "epoch": 0, "lr": 9.999999876242834e-05}
+{"train_loss": 3.809833526611328, "grad_norm": 0.2699662148952484, "global_step": 23, "epoch": 0, "lr": 9.999999876242834e-05}
+{"train_loss": 3.750337839126587, "grad_norm": 0.36923307180404663, "global_step": 24, "epoch": 0, "lr": 9.999999721546376e-05}
+{"train_loss": 3.7796430587768555, "grad_norm": 0.26082107424736023, "global_step": 25, "epoch": 0, "lr": 9.999999721546376e-05}
+{"train_loss": 3.7952685356140137, "grad_norm": 0.39062705636024475, "global_step": 26, "epoch": 0, "lr": 9.999999504971339e-05}
+{"train_loss": 3.7544047832489014, "grad_norm": 0.1965150386095047, "global_step": 27, "epoch": 0, "lr": 9.999999504971339e-05}
+{"train_loss": 3.7364883422851562, "grad_norm": 0.3700607120990753, "global_step": 28, "epoch": 0, "lr": 9.999999226517724e-05}
+{"train_loss": 3.682318925857544, "grad_norm": 0.13031569123268127, "global_step": 29, "epoch": 0, "lr": 9.999999226517724e-05}
+{"train_loss": 3.750153064727783, "grad_norm": 0.35022681951522827, "global_step": 30, "epoch": 0, "lr": 9.999998886185536e-05}
+83974776e-05}
+{"train_loss": 3.7710022926330566, "grad_norm": 1.2692668437957764, "global_step": 33, "epoch": 0, "lr": 9.999998483974776e-05}
+{"train_loss": 3.7969400882720947, "grad_norm": 2.4851691722869873, "global_step": 34, "epoch": 0, "lr": 9.999998019885453e-05}
+{"train_loss": 3.7792279720306396, "grad_norm": 2.512002468109131, "global_step": 35, "epoch": 0, "lr": 9.999998019885453e-05}
+{"train_loss": 3.8288862705230713, "grad_norm": 4.66420841217041, "global_step": 36, "epoch": 0, "lr": 9.99999749391757e-05}
+{"train_loss": 3.793712854385376, "grad_norm": 2.3314995765686035, "global_step": 37, "epoch": 0, "lr": 9.99999749391757e-05}
+{"train_loss": 3.8312277793884277, "grad_norm": 4.78786039352417, "global_step": 38, "epoch": 0, "lr": 9.999996906071134e-05}
+{"train_loss": 3.778836488723755, "grad_norm": 1.9080649614334106, "global_step": 39, "epoch": 0, "lr": 9.999996906071134e-05}
+{"train_loss": 3.802455186843872, "grad_norm": 3.4520087242126465, "global_step": 40, "epoch": 0, "lr": 9.999996256346153e-05}
+{"train_loss": 3.815281867980957, "grad_norm": 0.19909827411174774, "global_step": 41, "epoch": 0, "lr": 9.999996256346153e-05}
+{"train_loss": 3.768871545791626, "grad_norm": 0.5638014674186707, "global_step": 42, "epoch": 0, "lr": 9.999995544742635e-05}
+{"train_loss": 3.765110731124878, "grad_norm": 0.35555127263069153, "global_step": 43, "epoch": 0, "lr": 9.999995544742635e-05}
+{"train_loss": 3.7661640644073486, "grad_norm": 0.7404451370239258, "global_step": 44, "epoch": 0, "lr": 9.999994771260588e-05}
+{"train_loss": 3.762078285217285, "grad_norm": 0.3399713635444641, "global_step": 45, "epoch": 0, "lr": 9.999994771260588e-05}
+{"train_loss": 3.7468042373657227, "grad_norm": 0.7716804146766663, "global_step": 46, "epoch": 0, "lr": 9.999993935900023e-05}
+{"train_loss": 3.7671666145324707, "grad_norm": 0.1592184156179428, "global_step": 47, "epoch": 0, "lr": 9.999993935900023e-05}
+{"train_loss": 3.764308452606201, "grad_norm": 0.39841482043266296, "global_step": 48, "epoch": 0, "lr": 9.999993038660947e-05}
+{"train_loss": 3.766148567199707, "grad_norm": 0.34506940841674805, "global_step": 49, "epoch": 0, "lr": 9.999993038660947e-05}
+{"train_loss": 3.787818193435669, "grad_norm": 0.5775526762008667, "global_step": 50, "epoch": 0, "lr": 9.999992079543376e-05}
+{"train_loss": 3.7984201908111572, "grad_norm": 0.37691184878349304, "global_step": 51, "epoch": 0, "lr": 9.999992079543376e-05}
+{"train_loss": 3.7501060962677, "grad_norm": 0.7029170989990234, "global_step": 52, "epoch": 0, "lr": 9.999991058547318e-05}
+{"train_loss": 3.801084518432617, "grad_norm": 0.3741215765476227, "global_step": 53, "epoch": 0, "lr": 9.999991058547318e-05}
+{"train_loss": 3.7945199012756348, "grad_norm": 0.9044532775878906, "global_step": 54, "epoch": 0, "lr": 9.999989975672788e-05}
+{"train_loss": 3.77378511428833, "grad_norm": 0.19874800741672516, "global_step": 55, "epoch": 0, "lr": 9.999989975672788e-05}
+{"train_loss": 3.7798566818237305, "grad_norm": 0.45759376883506775, "global_step": 56, "epoch": 0, "lr": 9.999988830919797e-05}
+{"train_loss": 3.7658371925354004, "grad_norm": 1.1370397806167603, "global_step": 57, "epoch": 0, "lr": 9.999988830919797e-05}
+{"train_loss": 3.7611284255981445, "grad_norm": 2.2171638011932373, "global_step": 58, "epoch": 0, "lr": 9.999987624288363e-05}
+{"train_loss": 3.7753424644470215, "grad_norm": 1.6077601909637451, "global_step": 59, "epoch": 0, "lr": 9.999987624288363e-05}
+{"train_loss": 3.7917306423187256, "grad_norm": 3.2862675189971924, "global_step": 60, "epoch": 0, "lr": 9.999986355778496e-05}
+{"train_loss": 3.7183220386505127, "grad_norm": 1.6992913484573364, "global_step": 61, "epoch": 0, "lr": 9.999986355778496e-05}
+{"train_loss": 3.7725918292999268, "grad_norm": 3.287337303161621, "global_step": 62, "epoch": 0, "lr": 9.999985025390215e-05}
+{"train_loss": 3.771272897720337, "grad_norm": 1.37978994846344, "global_step": 63, "epoch": 0, "lr": 9.999985025390215e-05}
+{"train_loss": 3.730681896209717, "grad_norm": 2.5724732875823975, "global_step": 64, "epoch": 0, "lr": 9.999983633123536e-05}
+{"train_loss": 3.7703099250793457, "grad_norm": 0.22329014539718628, "global_step": 65, "epoch": 0, "lr": 9.999983633123536e-05}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/train.log

@@ -0,0 +1,12 @@
+[2026-01-28 00:19:04,613][numexpr.utils][INFO] - Note: detected 112 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2026-01-28 00:19:04,613][numexpr.utils][INFO] - Note: detected 112 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2026-01-28 00:19:04,613][numexpr.utils][INFO] - Note: NumExpr detected 112 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2026-01-28 00:19:04,613][numexpr.utils][INFO] - Note: NumExpr detected 112 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2026-01-28 00:19:04,613][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2026-01-28 00:19:04,613][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2026-01-28 00:19:10,581][datasets][INFO] - PyTorch version 2.2.2 available.
+[2026-01-28 00:19:10,581][datasets][INFO] - PyTorch version 2.2.2 available.
+[2026-01-28 00:19:10,582][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2026-01-28 00:19:10,582][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2026-01-28 00:19:10,583][datasets][INFO] - JAX version 0.4.30 available.
+[2026-01-28 00:19:10,583][datasets][INFO] - JAX version 0.4.30 available.

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/debug-internal.log

@@ -0,0 +1,9 @@
+{"time":"2026-01-28T00:19:15.378788051+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-28T00:19:15.37879851+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug-core.log"}
+{"time":"2026-01-28T00:19:15.50436033+08:00","level":"INFO","msg":"created new stream","id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.504391668+08:00","level":"INFO","msg":"stream: started","id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.504410866+08:00","level":"INFO","msg":"sender: started","stream_id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.504406808+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"zk1al3v3"}}
+{"time":"2026-01-28T00:19:15.504418481+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"zk1al3v3"}}
+{"time":"2026-01-28T00:19:16.166195204+08:00","level":"INFO","msg":"api: retrying HTTP error","status":409,"url":"https://api.wandb.ai/graphql"}
+{"time":"2026-01-28T00:19:18.738457616+08:00","level":"INFO","msg":"Starting system monitor"}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/debug.log

@@ -0,0 +1,26 @@
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Configure stats pid to 1113694
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [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'}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug.log
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug-internal.log
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():619] calling init triggers
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_mlp_projector', '_target_': 'llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace', '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_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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.19.02_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}, 'dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda', 'seed': 42, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 100, 'gradient_accumulate_every': 2, 'grad_norm_clip': 1, 'enable_normalizer': True, 'checkpoint_every': 1, 'val_every': 1, 'sample_every': 1, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'llm_module_training', 'resume': True, 'mode': 'online', 'name': '2026.01.28-00.19.02_train_mlp_projector_box-close-v2', 'tags': ['train_mlp_projector', 'box-close-v2', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'val_loss', 'mode': 'min', 'k': 5, 'format_str': 'epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2', 'wandb_name_base': '2026.01.28-00.19.02_train_mlp_projector_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-general-o4-mini.pt', 'data_path2': 'datasets/box-close-v2-o4-mini.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.05, 'dummy_normalizer': False}, '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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.19.02_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():669] starting backend
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():673] sending inform_init request
+2026-01-28 00:19:15,374 INFO    MainThread:1113694 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-28 00:19:15,374 INFO    MainThread:1113694 [wandb_init.py:init():686] backend started and connected
+2026-01-28 00:19:15,383 INFO    MainThread:1113694 [wandb_init.py:init():781] updated telemetry
+2026-01-28 00:19:15,442 INFO    MainThread:1113694 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-28 00:19:18,733 INFO    MainThread:1113694 [wandb_init.py:init():867] starting run threads in backend
+2026-01-28 00:19:18,925 INFO    MainThread:1113694 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-28 00:19:18,925 INFO    MainThread:1113694 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-28 00:19:18,926 INFO    MainThread:1113694 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-28 00:19:18,926 INFO    MainThread:1113694 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-28 00:19:18,928 INFO    MainThread:1113694 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-28 00:19:18,928 INFO    MainThread:1113694 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2'}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/files/output.log

@@ -0,0 +1 @@
+                                                                                              

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/files/requirements.txt

@@ -0,0 +1,282 @@
+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

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/files/wandb-metadata.json

@@ -0,0 +1,106 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.28",
+  "python": "3.9.25",
+  "startedAt": "2026-01-27T16:19:14.951887Z",
+  "args": [
+    "--config-path",
+    "config/lang_source",
+    "--config-name",
+    "llmbc_mlp_projector_box-close-v2_o4-mini.yaml"
+  ],
+  "program": "/work/u1131674/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "2cf60c4cdae3a2ac623eb2aebb7d12dc2beb2b66"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2",
+  "host": "hgpn21",
+  "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": "217153343488"
+    }
+  },
+  "memory": {
+    "total": "2163677220864"
+  },
+  "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:0x000000F000000F00000000000000",
+    "cpu_bind_list": "0x000000F000000F00000000000000",
+    "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": "1769703540",
+    "job_gid": "106773",
+    "job_group": "MST114558",
+    "job_id": "105042",
+    "job_name": "python3",
+    "job_nodelist": "hgpn21",
+    "job_num_nodes": "1",
+    "job_partition": "normal",
+    "job_qos": "normal",
+    "job_start_time": "1769530740",
+    "job_uid": "41408",
+    "job_user": "u1131674",
+    "jobid": "105042",
+    "launch_node_ipaddr": "172.21.101.1",
+    "localid": "0",
+    "mem_per_node": "204800",
+    "nnodes": "1",
+    "nodeid": "0",
+    "nodelist": "hgpn21",
+    "nprocs": "1",
+    "ntasks": "1",
+    "prio_process": "0",
+    "procid": "0",
+    "srun_comm_host": "172.21.101.1",
+    "srun_comm_port": "48065",
+    "step_gpus": "4",
+    "step_id": "0",
+    "step_launcher_port": "48065",
+    "step_nodelist": "hgpn21",
+    "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": "1113696",
+    "tasks_per_node": "1",
+    "topology_addr": "ibsw1.hgpn21",
+    "topology_addr_pattern": "switch.node",
+    "tres_per_task": "cpu:8",
+    "umask": "0022"
+  },
+  "cudaVersion": "12.4"
+}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/logs/debug-core.log

@@ -0,0 +1,14 @@
+{"time":"2026-01-28T00:19:14.330994308+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpi2mmsebx/port-1113696.txt","pid":1113696,"debug":false,"disable-analytics":false}
+{"time":"2026-01-28T00:19:14.331017077+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-28T00:19:14.331372236+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":1113696}
+{"time":"2026-01-28T00:19:14.331366446+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":39117,"Zone":""}}
+{"time":"2026-01-28T00:19:14.524618184+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:34736"}
+{"time":"2026-01-28T00:19:14.782379815+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpx5xwrxwi/port-1113694.txt","pid":1113694,"debug":false,"disable-analytics":false}
+{"time":"2026-01-28T00:19:14.782398135+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-28T00:19:14.782736352+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":1113694}
+{"time":"2026-01-28T00:19:14.782728767+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":38005,"Zone":""}}
+{"time":"2026-01-28T00:19:14.952416693+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"zk1al3v3","id":"127.0.0.1:34736"}
+{"time":"2026-01-28T00:19:14.977719823+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:53454"}
+{"time":"2026-01-28T00:19:15.070600831+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"zk1al3v3","id":"127.0.0.1:34736"}
+{"time":"2026-01-28T00:19:15.374994574+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"zk1al3v3","id":"127.0.0.1:53454"}
+{"time":"2026-01-28T00:19:15.504396406+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"zk1al3v3","id":"127.0.0.1:53454"}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/logs/debug-internal.log

@@ -0,0 +1,8 @@
+{"time":"2026-01-28T00:19:14.953645987+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-28T00:19:14.95365662+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/logs/debug-core.log"}
+{"time":"2026-01-28T00:19:15.070564922+08:00","level":"INFO","msg":"created new stream","id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.070594716+08:00","level":"INFO","msg":"stream: started","id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.070627235+08:00","level":"INFO","msg":"sender: started","stream_id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.070620892+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"zk1al3v3"}}
+{"time":"2026-01-28T00:19:15.07061686+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"zk1al3v3"}}
+{"time":"2026-01-28T00:19:16.15739395+08:00","level":"INFO","msg":"Starting system monitor"}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/logs/debug.log

@@ -0,0 +1,26 @@
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_setup.py:_flush():79] Configure stats pid to 1113696
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [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'}
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/logs/debug.log
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001914-zk1al3v3/logs/debug-internal.log
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_init.py:init():619] calling init triggers
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_mlp_projector', '_target_': 'llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace', '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_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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.19.02_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}, 'dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda', 'seed': 42, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 100, 'gradient_accumulate_every': 2, 'grad_norm_clip': 1, 'enable_normalizer': True, 'checkpoint_every': 1, 'val_every': 1, 'sample_every': 1, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'llm_module_training', 'resume': True, 'mode': 'online', 'name': '2026.01.28-00.19.02_train_mlp_projector_box-close-v2', 'tags': ['train_mlp_projector', 'box-close-v2', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'val_loss', 'mode': 'min', 'k': 5, 'format_str': 'epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2', 'wandb_name_base': '2026.01.28-00.19.02_train_mlp_projector_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-general-o4-mini.pt', 'data_path2': 'datasets/box-close-v2-o4-mini.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.05, '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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.19.02_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-28 00:19:14,949 INFO    MainThread:1113696 [wandb_init.py:init():669] starting backend
+2026-01-28 00:19:14,950 INFO    MainThread:1113696 [wandb_init.py:init():673] sending inform_init request
+2026-01-28 00:19:14,951 INFO    MainThread:1113696 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-28 00:19:14,951 INFO    MainThread:1113696 [wandb_init.py:init():686] backend started and connected
+2026-01-28 00:19:14,958 INFO    MainThread:1113696 [wandb_init.py:init():781] updated telemetry
+2026-01-28 00:19:15,013 INFO    MainThread:1113696 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-28 00:19:16,152 INFO    MainThread:1113696 [wandb_init.py:init():867] starting run threads in backend
+2026-01-28 00:19:16,391 INFO    MainThread:1113696 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-28 00:19:16,391 INFO    MainThread:1113696 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-28 00:19:16,391 INFO    MainThread:1113696 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-28 00:19:16,391 INFO    MainThread:1113696 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-28 00:19:16,394 INFO    MainThread:1113696 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-28 00:19:16,394 INFO    MainThread:1113696 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2'}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/files/output.log

@@ -0,0 +1,2 @@
+Training epoch 0:   0%|          | 0/565 [00:00<?, ?it/s]wandb: 409 encountered ({"errors":[{"message":"Error 1062 (23000): Duplicate entry '40535891-zk1al3v3' for key 'runs.PRIMARY'","path":["upsertBucket"]}],"data":{"upsertBucket":null}}), retrying request
+Training epoch 0:   5%|▌         | 30/565 [00:20<06:01,  1.48it/s, grad_norm=0.13, loss=3.68] 

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/files/requirements.txt

@@ -0,0 +1,282 @@
+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

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/files/wandb-metadata.json

@@ -0,0 +1,106 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.28",
+  "python": "3.9.25",
+  "startedAt": "2026-01-27T16:19:15.374492Z",
+  "args": [
+    "--config-path",
+    "config/lang_source",
+    "--config-name",
+    "llmdp_mlp_projector_box-close-v2_o4-mini.yaml"
+  ],
+  "program": "/work/u1131674/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "2cf60c4cdae3a2ac623eb2aebb7d12dc2beb2b66"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2",
+  "host": "hgpn21",
+  "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": "217153368064"
+    }
+  },
+  "memory": {
+    "total": "2163677220864"
+  },
+  "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:0x0000FF0000000000000000000000",
+    "cpu_bind_list": "0x0000FF0000000000000000000000",
+    "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": "1769703540",
+    "job_gid": "106773",
+    "job_group": "MST114558",
+    "job_id": "105045",
+    "job_name": "python3",
+    "job_nodelist": "hgpn21",
+    "job_num_nodes": "1",
+    "job_partition": "normal",
+    "job_qos": "normal",
+    "job_start_time": "1769530740",
+    "job_uid": "41408",
+    "job_user": "u1131674",
+    "jobid": "105045",
+    "launch_node_ipaddr": "172.21.101.1",
+    "localid": "0",
+    "mem_per_node": "204800",
+    "nnodes": "1",
+    "nodeid": "0",
+    "nodelist": "hgpn21",
+    "nprocs": "1",
+    "ntasks": "1",
+    "prio_process": "0",
+    "procid": "0",
+    "srun_comm_host": "172.21.101.1",
+    "srun_comm_port": "46533",
+    "step_gpus": "7",
+    "step_id": "0",
+    "step_launcher_port": "46533",
+    "step_nodelist": "hgpn21",
+    "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": "1113694",
+    "tasks_per_node": "1",
+    "topology_addr": "ibsw1.hgpn21",
+    "topology_addr_pattern": "switch.node",
+    "tres_per_task": "cpu:8",
+    "umask": "0022"
+  },
+  "cudaVersion": "12.4"
+}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug-core.log

@@ -0,0 +1,14 @@
+{"time":"2026-01-28T00:19:14.330994308+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpi2mmsebx/port-1113696.txt","pid":1113696,"debug":false,"disable-analytics":false}
+{"time":"2026-01-28T00:19:14.331017077+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-28T00:19:14.331372236+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":1113696}
+{"time":"2026-01-28T00:19:14.331366446+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":39117,"Zone":""}}
+{"time":"2026-01-28T00:19:14.524618184+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:34736"}
+{"time":"2026-01-28T00:19:14.782379815+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpx5xwrxwi/port-1113694.txt","pid":1113694,"debug":false,"disable-analytics":false}
+{"time":"2026-01-28T00:19:14.782398135+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-28T00:19:14.782736352+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":1113694}
+{"time":"2026-01-28T00:19:14.782728767+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":38005,"Zone":""}}
+{"time":"2026-01-28T00:19:14.952416693+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"zk1al3v3","id":"127.0.0.1:34736"}
+{"time":"2026-01-28T00:19:14.977719823+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:53454"}
+{"time":"2026-01-28T00:19:15.070600831+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"zk1al3v3","id":"127.0.0.1:34736"}
+{"time":"2026-01-28T00:19:15.374994574+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"zk1al3v3","id":"127.0.0.1:53454"}
+{"time":"2026-01-28T00:19:15.504396406+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"zk1al3v3","id":"127.0.0.1:53454"}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug-internal.log

@@ -0,0 +1,9 @@
+{"time":"2026-01-28T00:19:15.378788051+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-28T00:19:15.37879851+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug-core.log"}
+{"time":"2026-01-28T00:19:15.50436033+08:00","level":"INFO","msg":"created new stream","id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.504391668+08:00","level":"INFO","msg":"stream: started","id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.504410866+08:00","level":"INFO","msg":"sender: started","stream_id":"zk1al3v3"}
+{"time":"2026-01-28T00:19:15.504406808+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"zk1al3v3"}}
+{"time":"2026-01-28T00:19:15.504418481+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"zk1al3v3"}}
+{"time":"2026-01-28T00:19:16.166195204+08:00","level":"INFO","msg":"api: retrying HTTP error","status":409,"url":"https://api.wandb.ai/graphql"}
+{"time":"2026-01-28T00:19:18.738457616+08:00","level":"INFO","msg":"Starting system monitor"}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug.log

@@ -0,0 +1,26 @@
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Configure stats pid to 1113694
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-28 00:19:15,371 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [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'}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug.log
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/run-20260128_001915-zk1al3v3/logs/debug-internal.log
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():619] calling init triggers
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_mlp_projector', '_target_': 'llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace', '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_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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.19.02_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}, 'dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda', 'seed': 42, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 100, 'gradient_accumulate_every': 2, 'grad_norm_clip': 1, 'enable_normalizer': True, 'checkpoint_every': 1, 'val_every': 1, 'sample_every': 1, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'llm_module_training', 'resume': True, 'mode': 'online', 'name': '2026.01.28-00.19.02_train_mlp_projector_box-close-v2', 'tags': ['train_mlp_projector', 'box-close-v2', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'val_loss', 'mode': 'min', 'k': 5, 'format_str': 'epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2', 'wandb_name_base': '2026.01.28-00.19.02_train_mlp_projector_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-general-o4-mini.pt', 'data_path2': 'datasets/box-close-v2-o4-mini.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.05, 'dummy_normalizer': False}, '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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.19.02_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():669] starting backend
+2026-01-28 00:19:15,372 INFO    MainThread:1113694 [wandb_init.py:init():673] sending inform_init request
+2026-01-28 00:19:15,374 INFO    MainThread:1113694 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-28 00:19:15,374 INFO    MainThread:1113694 [wandb_init.py:init():686] backend started and connected
+2026-01-28 00:19:15,383 INFO    MainThread:1113694 [wandb_init.py:init():781] updated telemetry
+2026-01-28 00:19:15,442 INFO    MainThread:1113694 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-28 00:19:18,733 INFO    MainThread:1113694 [wandb_init.py:init():867] starting run threads in backend
+2026-01-28 00:19:18,925 INFO    MainThread:1113694 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-28 00:19:18,925 INFO    MainThread:1113694 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-28 00:19:18,926 INFO    MainThread:1113694 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-28 00:19:18,926 INFO    MainThread:1113694 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-28 00:19:18,928 INFO    MainThread:1113694 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-28 00:19:18,928 INFO    MainThread:1113694 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.19.02_train_mlp_projector_box-close-v2'}

2026.01.28/00.19.02_train_mlp_projector_box-close-v2/wandb/wandb-resume.json

@@ -0,0 +1 @@
+{"run_id": "zk1al3v3"}

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/.hydra/config.yaml

@@ -0,0 +1,138 @@
+name: train_mlp_projector
+_target_: llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace
+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_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}
+dataloader:
+  batch_size: 128
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 128
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.Adam
+  lr: 0.0001
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda
+  seed: 42
+  debug: false
+  resume: true
+  lr_scheduler: cosine
+  lr_warmup_steps: 10
+  num_epochs: 100
+  gradient_accumulate_every: 2
+  grad_norm_clip: 1
+  enable_normalizer: true
+  checkpoint_every: 1
+  val_every: 1
+  sample_every: 1
+  sample_max_batch: 128
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: llm_module_training
+  resume: true
+  mode: online
+  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+  tags:
+  - ${name}
+  - ${task_name}
+  - ${exp_name}
+  id: null
+  group: null
+checkpoint:
+  topk:
+    monitor_key: val_loss
+    mode: min
+    k: 5
+    format_str: epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt
+  save_last_ckpt: true
+  save_last_snapshot: false
+multi_run:
+  run_dir: data/outputs/${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-general-o4-mini.pt
+    data_path2: datasets/box-close-v2-o4-mini.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: 0.05
+    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: pretrained
+  finetune_mode: orig
+  checkpoint: ''
+  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}

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/.hydra/hydra.yaml

@@ -0,0 +1,154 @@
+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: []
+  job:
+    name: train
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: llmbc_mlp_projector_box-close-v2_o4-mini.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/lang_source
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_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

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/checkpoints/epoch=0093-val_loss=2.191.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c2e0a98bb77bf92c36e628ab66933ebb6a0f5c409800890b9dc261541581766
+size 4110822

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/checkpoints/epoch=0096-val_loss=2.189.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d30704ee3eae4bbb3d76f86e632315fc2c74f62a7fd3488744bb1cd5bd5a83b
+size 4110822

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/checkpoints/epoch=0097-val_loss=2.189.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a269c54f1cb469981f7d47c05e5c94945d5bea4c549e2463b9cbf4af9c8382bf
+size 4110822

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/checkpoints/epoch=0098-val_loss=2.186.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d87b6064f6b6d606df21820412b92e036664f0e8b8497d93480cf02776340180
+size 4110822

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/checkpoints/epoch=0099-val_loss=2.190.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f020bfaf26ab3446ecac695500c3c256cbfdf28541876f8858216f7ed58a083c
+size 4110822

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/checkpoints/latest.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f020bfaf26ab3446ecac695500c3c256cbfdf28541876f8858216f7ed58a083c
+size 4110822

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/train.log

@@ -0,0 +1,6 @@
+[2026-01-28 00:20:33,669][numexpr.utils][INFO] - Note: detected 112 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2026-01-28 00:20:33,669][numexpr.utils][INFO] - Note: NumExpr detected 112 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2026-01-28 00:20:33,669][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2026-01-28 00:20:38,172][datasets][INFO] - PyTorch version 2.2.2 available.
+[2026-01-28 00:20:38,173][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2026-01-28 00:20:38,174][datasets][INFO] - JAX version 0.4.30 available.

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/debug-internal.log

@@ -0,0 +1,20 @@
+{"time":"2026-01-28T00:20:42.287152799+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-28T00:20:42.287163685+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug-core.log"}
+{"time":"2026-01-28T00:20:42.40159487+08:00","level":"INFO","msg":"created new stream","id":"d9kt1a6y"}
+{"time":"2026-01-28T00:20:42.401631969+08:00","level":"INFO","msg":"stream: started","id":"d9kt1a6y"}
+{"time":"2026-01-28T00:20:42.40165535+08:00","level":"INFO","msg":"sender: started","stream_id":"d9kt1a6y"}
+{"time":"2026-01-28T00:20:42.401645639+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T00:20:42.401655093+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T00:20:43.122350505+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-01-28T03:18:28.843132355+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/files/chyang25-national-taiwan-university/llm_module_training/d9kt1a6y/file_stream"}
+{"time":"2026-01-28T04:22:17.072983497+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/files/chyang25-national-taiwan-university/llm_module_training/d9kt1a6y/file_stream"}
+{"time":"2026-01-28T05:23:28.602833473+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/graphql"}
+{"time":"2026-01-28T09:24:07.706274513+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/graphql"}
+{"time":"2026-01-28T10:52:44.707921101+08:00","level":"INFO","msg":"stream: closing","id":"d9kt1a6y"}
+{"time":"2026-01-28T10:52:44.707946286+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-01-28T10:52:44.708452053+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-01-28T10:52:48.257346098+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-01-28T10:52:48.506294252+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T10:52:48.506322189+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T10:52:48.506353262+08:00","level":"INFO","msg":"sender: closed","stream_id":"d9kt1a6y"}
+{"time":"2026-01-28T10:52:48.506389957+08:00","level":"INFO","msg":"stream: closed","id":"d9kt1a6y"}

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/debug.log

@@ -0,0 +1,27 @@
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Configure stats pid to 1129901
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [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'}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug.log
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug-internal.log
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():619] calling init triggers
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_mlp_projector', '_target_': 'llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace', '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_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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.20.31_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}, 'dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda', 'seed': 42, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 100, 'gradient_accumulate_every': 2, 'grad_norm_clip': 1, 'enable_normalizer': True, 'checkpoint_every': 1, 'val_every': 1, 'sample_every': 1, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'llm_module_training', 'resume': True, 'mode': 'online', 'name': '2026.01.28-00.20.31_train_mlp_projector_box-close-v2', 'tags': ['train_mlp_projector', 'box-close-v2', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'val_loss', 'mode': 'min', 'k': 5, 'format_str': 'epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2', 'wandb_name_base': '2026.01.28-00.20.31_train_mlp_projector_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-general-o4-mini.pt', 'data_path2': 'datasets/box-close-v2-o4-mini.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.05, '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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.20.31_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():669] starting backend
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():673] sending inform_init request
+2026-01-28 00:20:42,284 INFO    MainThread:1129901 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-28 00:20:42,285 INFO    MainThread:1129901 [wandb_init.py:init():686] backend started and connected
+2026-01-28 00:20:42,292 INFO    MainThread:1129901 [wandb_init.py:init():781] updated telemetry
+2026-01-28 00:20:42,346 INFO    MainThread:1129901 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-28 00:20:43,117 INFO    MainThread:1129901 [wandb_init.py:init():867] starting run threads in backend
+2026-01-28 00:20:43,315 INFO    MainThread:1129901 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-28 00:20:43,316 INFO    MainThread:1129901 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-28 00:20:43,316 INFO    MainThread:1129901 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-28 00:20:43,316 INFO    MainThread:1129901 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-28 00:20:43,319 INFO    MainThread:1129901 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-28 00:20:43,319 INFO    MainThread:1129901 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2'}
+2026-01-28 10:52:44,708 WARNING MsgRouterThr:1129901 [router.py:message_loop():75] message_loop has been closed

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/files/config.yaml

@@ -0,0 +1,242 @@
+_target_:
+    value: llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace
+_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
+                - 105
+            "2":
+                - 1
+                - 2
+                - 3
+                - 5
+                - 11
+                - 12
+                - 41
+                - 49
+                - 50
+                - 51
+                - 53
+                - 55
+                - 71
+                - 83
+                - 95
+                - 98
+                - 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}-val_loss={val_loss:.3f}.ckpt
+            k: 5
+            mode: min
+            monitor_key: val_loss
+dataloader:
+    value:
+        batch_size: 128
+        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: ""
+        config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+        finetune_mode: orig
+        hydra:
+            job:
+                override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+            run:
+                dir: data/outputs/2026.01.28/00.20.31_HuggingFaceTB/SmolLM2-135M-Instruct
+        llm_mode: pretrained
+        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_translator:
+    value:
+        _target_: llmbc.translator.llm_translator.LLMTranslator
+        action_dim: 4
+        cfg:
+            causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+            checkpoint: ""
+            config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+            finetune_mode: orig
+            hydra:
+                job:
+                    override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+                run:
+                    dir: data/outputs/2026.01.28/00.20.31_HuggingFaceTB/SmolLM2-135M-Instruct
+            llm_mode: pretrained
+            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
+logging:
+    value:
+        group: null
+        id: null
+        mode: online
+        name: 2026.01.28-00.20.31_train_mlp_projector_box-close-v2
+        project: llm_module_training
+        resume: true
+        tags:
+            - train_mlp_projector
+            - box-close-v2
+            - default
+model_name:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+multi_run:
+    value:
+        run_dir: data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2
+        wandb_name_base: 2026.01.28-00.20.31_train_mlp_projector_box-close-v2
+n_action_steps:
+    value: 1
+n_latency_steps:
+    value: 0
+n_obs_steps:
+    value: 1
+name:
+    value: train_mlp_projector
+obs_dim:
+    value: 9
+optimizer:
+    value:
+        _target_: torch.optim.Adam
+        betas:
+            - 0.95
+            - 0.999
+        eps: 1e-08
+        lr: 0.0001
+        weight_decay: 1e-06
+output_dir:
+    value: /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2
+past_action_visible:
+    value: false
+task:
+    value:
+        action_dim: 4
+        dataset:
+            _target_: llmbc.dataset.metaworld_lowdim_dataset.MetaworldLowdimDataset
+            data_path: datasets/box-close-v2-general-o4-mini.pt
+            data_path2: datasets/box-close-v2-o4-mini.pt
+            dummy_normalizer: true
+            horizon: 1
+            obs_eef_target: true
+            pad_after: 0
+            pad_before: 0
+            use_manual_normalizer: false
+            val_ratio: 0.05
+        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: 1
+        debug: false
+        device: cuda
+        enable_normalizer: true
+        grad_norm_clip: 1
+        gradient_accumulate_every: 2
+        lr_scheduler: cosine
+        lr_warmup_steps: 10
+        max_train_steps: null
+        max_val_steps: null
+        num_epochs: 100
+        resume: true
+        sample_every: 1
+        sample_max_batch: 128
+        seed: 42
+        tqdm_interval_sec: 1
+        val_every: 1
+val_dataloader:
+    value:
+        batch_size: 128
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: true

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/files/output.log

@@ -0,0 +1 @@
+                                                                                                

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/files/wandb-metadata.json

@@ -0,0 +1,106 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.28",
+  "python": "3.9.25",
+  "startedAt": "2026-01-27T16:20:42.285279Z",
+  "args": [
+    "--config-path",
+    "config/lang_source",
+    "--config-name",
+    "llmbc_mlp_projector_box-close-v2_o4-mini.yaml"
+  ],
+  "program": "/work/u1131674/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "2cf60c4cdae3a2ac623eb2aebb7d12dc2beb2b66"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2",
+  "host": "hgpn21",
+  "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": "217170878464"
+    }
+  },
+  "memory": {
+    "total": "2163677220864"
+  },
+  "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:0x000000F000000F00000000000000",
+    "cpu_bind_list": "0x000000F000000F00000000000000",
+    "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": "1769703630",
+    "job_gid": "106773",
+    "job_group": "MST114558",
+    "job_id": "105048",
+    "job_name": "python3",
+    "job_nodelist": "hgpn21",
+    "job_num_nodes": "1",
+    "job_partition": "normal",
+    "job_qos": "normal",
+    "job_start_time": "1769530830",
+    "job_uid": "41408",
+    "job_user": "u1131674",
+    "jobid": "105048",
+    "launch_node_ipaddr": "172.21.101.1",
+    "localid": "0",
+    "mem_per_node": "204800",
+    "nnodes": "1",
+    "nodeid": "0",
+    "nodelist": "hgpn21",
+    "nprocs": "1",
+    "ntasks": "1",
+    "prio_process": "0",
+    "procid": "0",
+    "srun_comm_host": "172.21.101.1",
+    "srun_comm_port": "41331",
+    "step_gpus": "4",
+    "step_id": "0",
+    "step_launcher_port": "41331",
+    "step_nodelist": "hgpn21",
+    "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": "1129901",
+    "tasks_per_node": "1",
+    "topology_addr": "ibsw1.hgpn21",
+    "topology_addr_pattern": "switch.node",
+    "tres_per_task": "cpu:8",
+    "umask": "0022"
+  },
+  "cudaVersion": "12.4"
+}

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/files/wandb-summary.json

@@ -0,0 +1 @@
+{"train_loss":2.2015625126593936,"_wandb":{"runtime":37922},"_step":56499,"_timestamp":1.7695687646872494e+09,"grad_norm":5.658727169036865,"global_step":56499,"epoch":99,"val_loss":2.1897988319396973,"lr":0,"_runtime":37922.42265669}

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug-core.log

@@ -0,0 +1,14 @@
+{"time":"2026-01-28T00:20:41.678122741+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpf51s48j1/port-1129901.txt","pid":1129901,"debug":false,"disable-analytics":false}
+{"time":"2026-01-28T00:20:41.678144133+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-01-28T00:20:41.678525998+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":1129901}
+{"time":"2026-01-28T00:20:41.678522018+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":41121,"Zone":""}}
+{"time":"2026-01-28T00:20:41.872810103+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T00:20:42.285792202+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"d9kt1a6y","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T00:20:42.401637789+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"d9kt1a6y","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T10:52:44.707861963+08:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T10:52:44.707928482+08:00","level":"INFO","msg":"server is shutting down"}
+{"time":"2026-01-28T10:52:44.707920412+08:00","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T10:52:44.708005087+08:00","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T10:52:48.506700484+08:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T10:52:48.50671573+08:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:33780"}
+{"time":"2026-01-28T10:52:48.506722098+08:00","level":"INFO","msg":"server is closed"}

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug-internal.log

@@ -0,0 +1,20 @@
+{"time":"2026-01-28T00:20:42.287152799+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-01-28T00:20:42.287163685+08:00","level":"INFO","msg":"created symlink","path":"/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug-core.log"}
+{"time":"2026-01-28T00:20:42.40159487+08:00","level":"INFO","msg":"created new stream","id":"d9kt1a6y"}
+{"time":"2026-01-28T00:20:42.401631969+08:00","level":"INFO","msg":"stream: started","id":"d9kt1a6y"}
+{"time":"2026-01-28T00:20:42.40165535+08:00","level":"INFO","msg":"sender: started","stream_id":"d9kt1a6y"}
+{"time":"2026-01-28T00:20:42.401645639+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T00:20:42.401655093+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T00:20:43.122350505+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-01-28T03:18:28.843132355+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/files/chyang25-national-taiwan-university/llm_module_training/d9kt1a6y/file_stream"}
+{"time":"2026-01-28T04:22:17.072983497+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/files/chyang25-national-taiwan-university/llm_module_training/d9kt1a6y/file_stream"}
+{"time":"2026-01-28T05:23:28.602833473+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/graphql"}
+{"time":"2026-01-28T09:24:07.706274513+08:00","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/graphql"}
+{"time":"2026-01-28T10:52:44.707921101+08:00","level":"INFO","msg":"stream: closing","id":"d9kt1a6y"}
+{"time":"2026-01-28T10:52:44.707946286+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-01-28T10:52:44.708452053+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-01-28T10:52:48.257346098+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-01-28T10:52:48.506294252+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T10:52:48.506322189+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"d9kt1a6y"}}
+{"time":"2026-01-28T10:52:48.506353262+08:00","level":"INFO","msg":"sender: closed","stream_id":"d9kt1a6y"}
+{"time":"2026-01-28T10:52:48.506389957+08:00","level":"INFO","msg":"stream: closed","id":"d9kt1a6y"}

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug.log

@@ -0,0 +1,27 @@
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Configure stats pid to 1129901
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Loading settings from /home/u1131674/.config/wandb/settings
+2026-01-28 00:20:42,282 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Loading settings from /work/u1131674/LLM-BC/wandb/settings
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [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'}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:_log_setup():533] Logging user logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug.log
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:_log_setup():534] Logging internal logs to /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/logs/debug-internal.log
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():619] calling init triggers
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_mlp_projector', '_target_': 'llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace', '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_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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.20.31_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 9, 'action_dim': 4, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}, 'dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 128, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.Adam', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda', 'seed': 42, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 10, 'num_epochs': 100, 'gradient_accumulate_every': 2, 'grad_norm_clip': 1, 'enable_normalizer': True, 'checkpoint_every': 1, 'val_every': 1, 'sample_every': 1, 'sample_max_batch': 128, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'llm_module_training', 'resume': True, 'mode': 'online', 'name': '2026.01.28-00.20.31_train_mlp_projector_box-close-v2', 'tags': ['train_mlp_projector', 'box-close-v2', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'val_loss', 'mode': 'min', 'k': 5, 'format_str': 'epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2', 'wandb_name_base': '2026.01.28-00.20.31_train_mlp_projector_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-general-o4-mini.pt', 'data_path2': 'datasets/box-close-v2-o4-mini.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.05, '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': 'pretrained', 'finetune_mode': 'orig', 'checkpoint': '', '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/2026.01.28/00.20.31_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():669] starting backend
+2026-01-28 00:20:42,283 INFO    MainThread:1129901 [wandb_init.py:init():673] sending inform_init request
+2026-01-28 00:20:42,284 INFO    MainThread:1129901 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-01-28 00:20:42,285 INFO    MainThread:1129901 [wandb_init.py:init():686] backend started and connected
+2026-01-28 00:20:42,292 INFO    MainThread:1129901 [wandb_init.py:init():781] updated telemetry
+2026-01-28 00:20:42,346 INFO    MainThread:1129901 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-01-28 00:20:43,117 INFO    MainThread:1129901 [wandb_init.py:init():867] starting run threads in backend
+2026-01-28 00:20:43,315 INFO    MainThread:1129901 [wandb_run.py:_console_start():2451] atexit reg
+2026-01-28 00:20:43,316 INFO    MainThread:1129901 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-01-28 00:20:43,316 INFO    MainThread:1129901 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-01-28 00:20:43,316 INFO    MainThread:1129901 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-01-28 00:20:43,319 INFO    MainThread:1129901 [wandb_init.py:init():911] run started, returning control to user process
+2026-01-28 00:20:43,319 INFO    MainThread:1129901 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/work/u1131674/LLM-BC/data/outputs/2026.01.28/00.20.31_train_mlp_projector_box-close-v2'}
+2026-01-28 10:52:44,708 WARNING MsgRouterThr:1129901 [router.py:message_loop():75] message_loop has been closed

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/run-20260128_002042-d9kt1a6y/run-d9kt1a6y.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de7f3a5e581eb8e1ef74351546fe00ccfbe437c9826d4574fddbca5ecf91d3df
+size 55111506

2026.01.28/00.20.31_train_mlp_projector_box-close-v2/wandb/wandb-resume.json

@@ -0,0 +1 @@
+{"run_id": "d9kt1a6y"}

2026.01.28/00.21.32_train_mlp_projector_box-close-v2/.hydra/config.yaml

@@ -0,0 +1,138 @@
+name: train_mlp_projector
+_target_: llmbc.workspace.train_mlp_projector_workspace.TrainMlpProjectorWorkspace
+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_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}
+dataloader:
+  batch_size: 128
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 128
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.Adam
+  lr: 0.0001
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda
+  seed: 42
+  debug: false
+  resume: true
+  lr_scheduler: cosine
+  lr_warmup_steps: 10
+  num_epochs: 100
+  gradient_accumulate_every: 2
+  grad_norm_clip: 1
+  enable_normalizer: true
+  checkpoint_every: 1
+  val_every: 1
+  sample_every: 1
+  sample_max_batch: 128
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: llm_module_training
+  resume: true
+  mode: online
+  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+  tags:
+  - ${name}
+  - ${task_name}
+  - ${exp_name}
+  id: null
+  group: null
+checkpoint:
+  topk:
+    monitor_key: val_loss
+    mode: min
+    k: 5
+    format_str: epoch={epoch:04d}-val_loss={val_loss:.3f}.ckpt
+  save_last_ckpt: true
+  save_last_snapshot: false
+multi_run:
+  run_dir: data/outputs/${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-general-o4-mini.pt
+    data_path2: datasets/box-close-v2-o4-mini.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: 0.05
+    dummy_normalizer: false
+  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: pretrained
+  finetune_mode: orig
+  checkpoint: ''
+  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}

2026.01.28/00.21.32_train_mlp_projector_box-close-v2/.hydra/hydra.yaml

@@ -0,0 +1,154 @@
+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: []
+  job:
+    name: train
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: llmdp_mlp_projector_box-close-v2_o4-mini.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/lang_source
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /work/u1131674/LLM-BC/data/outputs/2026.01.28/00.21.32_train_mlp_projector_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

2026.01.28/00.21.32_train_mlp_projector_box-close-v2/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

2026.01.28/00.21.32_train_mlp_projector_box-close-v2/checkpoints/epoch=0084-val_loss=2.250.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4d952071a7dfa64895a14c7448ff390dfe48e6e1b386bc7683150d87dd8bcba
+size 4114690

2026.01.28/00.21.32_train_mlp_projector_box-close-v2/checkpoints/epoch=0093-val_loss=2.251.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc9b8dc898db74eaaad6ecde020091df9f51d8f80ce39af5646ef50a72c4f90e
+size 4114690