Upload folder using huggingface_hub

.gitattributes

@@ -39,3 +39,11 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 2025.12.17/06.16.16_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/logs.json.txt filter=lfs diff=lfs merge=lfs -text
 2025.12.17/06.16.16_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251217_061628-ktb7lf93/run-ktb7lf93.wandb filter=lfs diff=lfs merge=lfs -text
 2025.12.17/06.18.45_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_061857-x4c1ls8y/run-x4c1ls8y.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/run-t2ymaznj.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/run-cflo0geg.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172356-7wg8nalk/run-7wg8nalk.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.38_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2025.12.20/17.23.38_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172400-9iiqcyau/run-9iiqcyau.wandb filter=lfs diff=lfs merge=lfs -text

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/config.yaml

@@ -0,0 +1,193 @@
+name: train_llm_diffusion_unet_lowdim
+_target_: llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace
+obs_dim: ${task.obs_dim}
+action_dim: ${task.action_dim}
+task_name: ${task.name}
+exp_name: default
+model_name: ${llm.name}
+horizon: 1
+n_obs_steps: 1
+n_action_steps: 1
+n_latency_steps: 0
+past_action_visible: false
+keypoint_visible_rate: 1.0
+obs_as_local_cond: false
+obs_as_global_cond: true
+pred_action_steps_only: false
+llm_orig_expert_feedback: true
+llm_do_sample: false
+policy:
+  _target_: llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy
+  model:
+    _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
+    input_dim: '${eval: ${task.action_dim} if ${obs_as_local_cond} or ${obs_as_global_cond}
+      else ${task.obs_dim} + ${task.action_dim}}'
+    local_cond_dim: '${eval: ${task.obs_dim} if ${obs_as_local_cond} else None}'
+    global_cond_dim: '${eval: ${task.obs_dim}*${n_obs_steps} if ${obs_as_global_cond}
+      else None}'
+    diffusion_step_embed_dim: 256
+    down_dims:
+    - 256
+    - 512
+    - 1024
+    kernel_size: 5
+    n_groups: 8
+    cond_predict_scale: true
+  noise_scheduler:
+    _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
+    num_train_timesteps: 100
+    beta_start: 0.0001
+    beta_end: 0.02
+    beta_schedule: squaredcos_cap_v2
+    variance_type: fixed_small
+    clip_sample: true
+    prediction_type: epsilon
+  horizon: ${horizon}
+  obs_dim: ${obs_dim}
+  action_dim: ${action_dim}
+  n_action_steps: ${eval:'${n_action_steps}+${n_latency_steps}'}
+  n_obs_steps: ${n_obs_steps}
+  num_inference_steps: 100
+  obs_as_local_cond: ${obs_as_local_cond}
+  obs_as_global_cond: ${obs_as_global_cond}
+  pred_action_steps_only: ${pred_action_steps_only}
+  oa_step_convention: true
+  llm_discriminator:
+    _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+    task_id: ${task_name}
+    llm_translator:
+      _target_: llmbc.translator.llm_translator.LLMTranslator
+      cfg: ${llm}
+      obs_dim: ${obs_dim}
+      action_dim: ${action_dim}
+      horizon: ${horizon}
+      n_obs_steps: ${n_obs_steps}
+      n_action_steps: ${eval:'${n_action_steps}+${n_latency_steps}'}
+  loss_dp_weight: 1.0
+  loss_llm_weight: 0.0001
+  normalize_llm_loss: true
+  reweight_llm_loss: true
+ema:
+  _target_: llmbc.model.diffusion.ema_model.EMAModel
+  update_after_step: 0
+  inv_gamma: 1.0
+  power: 0.75
+  min_value: 0.0
+  max_value: 0.9999
+dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: false
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.AdamW
+  lr: 0.0001
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda:0
+  seed: 43
+  debug: false
+  resume: true
+  lr_scheduler: cosine
+  lr_warmup_steps: 500
+  num_epochs: 1001
+  gradient_accumulate_every: 16
+  use_ema: true
+  rollout_every: 5
+  checkpoint_every: 5
+  val_every: 1
+  sample_every: 5
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: ${task.name}-training
+  resume: true
+  mode: online
+  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+  tags:
+  - ${name}
+  - ${task_name}
+  - ${exp_name}
+  id: null
+  group: null
+checkpoint:
+  topk:
+    monitor_key: test_success_rate
+    mode: max
+    k: 5
+    format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
+  save_last_ckpt: true
+  save_last_snapshot: false
+multi_run:
+  run_dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  wandb_name_base: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+task:
+  name: maze2d-medium-v0
+  obs_dim: 6
+  action_dim: 2
+  env_runner:
+    _target_: llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner
+    env_name: llf-pointmaze-maze2d-medium-v0
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 250
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    - fp
+    visual: false
+    discount: 0.99
+  dataset:
+    _target_: llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset
+    data_path: datasets/maze2d-medium-v0.pt
+    data_path2: datasets/maze2d-medium-v0.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: 0.02
+    dummy_normalizer: false
+  instructor:
+    _target_: llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor
+llm:
+  name: HuggingFaceTB/SmolLM2-135M-Instruct
+  model_name: SmolLM2-135M-Instruct
+  config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+  causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+  use_quantization: false
+  use_joint_mlp_projector: true
+  llm_mode: ete-finetuned
+  finetune_mode: orig
+  checkpoint: data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850
+  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}

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/hydra.yaml

@@ -0,0 +1,156 @@
+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:
+    - policy.loss_llm_weight=1.0e-4
+    - training.seed=43
+  job:
+    name: train
+    chdir: null
+    override_dirname: policy.loss_llm_weight=1.0e-4,training.seed=43
+    id: ???
+    num: ???
+    config_name: llmdp_maze2d-medium-v0.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+    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

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/overrides.yaml

@@ -0,0 +1,2 @@
+- policy.loss_llm_weight=1.0e-4
+- training.seed=43

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0530-test_success_rate=0.940.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edb15e65dae3f2d901b98fa3d66b04062d15be29a5400893710322911e12f37e
+size 1039753586

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0645-test_success_rate=0.940.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:470e7f84ae31e6d3b45d339b21862ec984965b56b4164df8dc2f8eb38ecfd051
+size 1039753586

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0790-test_success_rate=0.940.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a4e527287d83f76bbfd45184d67ad12d0a0a4ed36edbef1ca9fa2da36cf59ce
+size 1039753586

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0800-test_success_rate=0.940.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3ab310f9ef03b29594786d218db2729b7557a6f5a44d66626e2f7284d47d4a1
+size 1039753586

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0840-test_success_rate=0.940.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f81ea3770b2d974ed0e7f93682717dcc3b4542eb4dae1dd5962ef7d6532b21cd
+size 1039753586

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/latest.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1996d47db98030ee3431f3c1cd6f49b71ae5f7981ae36a878b5296c861e58358
+size 1039753586

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/logs.json.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99cce0b3ab9d1543151be3003e1e2c68a96375d91cefd11e12fde8d602153daa
+size 278801978

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/train.log

@@ -0,0 +1,9 @@
+[2025-12-20 17:23:20,380][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2025-12-20 17:23:20,380][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2025-12-20 17:23:20,380][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2025-12-20 17:23:22,519][datasets][INFO] - PyTorch version 2.2.2 available.
+[2025-12-20 17:23:22,520][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2025-12-20 17:23:22,521][datasets][INFO] - JAX version 0.4.30 available.
+[2025-12-20 17:23:24,418][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.496819e+07
+[2025-12-20 17:23:39,358][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
+[2025-12-20 17:23:39,364][absl][INFO] - MuJoCo library version is: 2.3.7

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/debug-internal.log

@@ -0,0 +1,28 @@
+{"time":"2025-12-20T17:23:40.39265505Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-20T17:23:40.392690291Z","level":"INFO","msg":"created symlink","path":"/workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug-core.log"}
+{"time":"2025-12-20T17:23:40.50006116Z","level":"INFO","msg":"created new stream","id":"t2ymaznj"}
+{"time":"2025-12-20T17:23:40.500106816Z","level":"INFO","msg":"stream: started","id":"t2ymaznj"}
+{"time":"2025-12-20T17:23:40.500367774Z","level":"INFO","msg":"sender: started","stream_id":"t2ymaznj"}
+{"time":"2025-12-20T17:23:40.500120299Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-20T17:23:40.500143338Z","level":"INFO","msg":"handler: started","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-20T17:23:41.157625473Z","level":"INFO","msg":"Starting system monitor"}
+{"time":"2025-12-22T02:02:44.208411298Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:02:59.788100106Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": read tcp 192.168.19.170:48608->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-22T05:06:59.165164769Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T05:09:29.278598582Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T06:16:10.957017478Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T06:17:26.36373148Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:24:33.296938897Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:24:44.346922515Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:16.775885945Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:51.300218869Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:28:59.360396234Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:32:46.734796863Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": read tcp 192.168.19.170:44740->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-27T20:10:03.581119512Z","level":"INFO","msg":"stream: closing","id":"t2ymaznj"}
+{"time":"2025-12-27T20:10:03.581183738Z","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2025-12-27T20:10:03.583096625Z","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2025-12-27T20:10:04.923009029Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2025-12-27T20:10:05.183243732Z","level":"INFO","msg":"handler: closed","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-27T20:10:05.183318468Z","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-27T20:10:05.183326493Z","level":"INFO","msg":"sender: closed","stream_id":"t2ymaznj"}
+{"time":"2025-12-27T20:10:05.18352636Z","level":"INFO","msg":"stream: closed","id":"t2ymaznj"}

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/debug.log

@@ -0,0 +1,27 @@
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Configure stats pid to 14556
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Loading settings from /workspace/LLM-BC/wandb/settings
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/workspace/LLM-BC/train.py', 'program': '/workspace/LLM-BC/./train.py'}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:_log_setup():533] Logging user logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug.log
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:_log_setup():534] Logging internal logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug-internal.log
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:init():619] calling init triggers
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llm_diffusion_unet_lowdim', '_target_': 'llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace', 'obs_dim': 6, 'action_dim': 2, 'task_name': 'maze2d-medium-v0', '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, 'keypoint_visible_rate': 1.0, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 2, 'local_cond_dim': None, 'global_cond_dim': 6, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'horizon': 1, 'obs_dim': 6, 'action_dim': 2, 'n_action_steps': 1, 'n_obs_steps': 1, 'num_inference_steps': 100, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'oa_step_convention': True, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'maze2d-medium-v0', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.18_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 6, 'action_dim': 2, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.0001, 'normalize_llm_loss': True, 'reweight_llm_loss': True}, 'ema': {'_target_': 'llmbc.model.diffusion.ema_model.EMAModel', 'update_after_step': 0, 'inv_gamma': 1.0, 'power': 0.75, 'min_value': 0.0, 'max_value': 0.9999}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': False, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 43, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'num_epochs': 1001, 'gradient_accumulate_every': 16, 'use_ema': True, 'rollout_every': 5, 'checkpoint_every': 5, 'val_every': 1, 'sample_every': 5, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'maze2d-medium-v0-training', 'resume': True, 'mode': 'online', 'name': '2025.12.20-17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'tags': ['train_llm_diffusion_unet_lowdim', 'maze2d-medium-v0', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'wandb_name_base': '2025.12.20-17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}, 'task': {'name': 'maze2d-medium-v0', 'obs_dim': 6, 'action_dim': 2, 'env_runner': {'_target_': 'llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner', 'env_name': 'llf-pointmaze-maze2d-medium-v0', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 250, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset', 'data_path': 'datasets/maze2d-medium-v0.pt', 'data_path2': 'datasets/maze2d-medium-v0.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.18_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-20 17:23:40,389 INFO    MainThread:14556 [wandb_init.py:init():669] starting backend
+2025-12-20 17:23:40,389 INFO    MainThread:14556 [wandb_init.py:init():673] sending inform_init request
+2025-12-20 17:23:40,390 INFO    MainThread:14556 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-20 17:23:40,390 INFO    MainThread:14556 [wandb_init.py:init():686] backend started and connected
+2025-12-20 17:23:40,397 INFO    MainThread:14556 [wandb_init.py:init():781] updated telemetry
+2025-12-20 17:23:40,470 INFO    MainThread:14556 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-20 17:23:41,128 INFO    MainThread:14556 [wandb_init.py:init():867] starting run threads in backend
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-20 17:23:42,624 INFO    MainThread:14556 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-20 17:23:42,624 INFO    MainThread:14556 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}
+2025-12-27 20:10:03,581 WARNING MsgRouterThr:14556 [router.py:message_loop():75] message_loop has been closed

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/files/config.yaml

@@ -0,0 +1,304 @@
+_target_:
+    value: llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace
+_wandb:
+    value:
+        cli_version: 0.18.6
+        m: []
+        python_version: 3.9.25
+        t:
+            "1":
+                - 1
+                - 2
+                - 3
+                - 5
+                - 11
+                - 12
+                - 41
+                - 49
+                - 50
+                - 51
+                - 53
+                - 55
+                - 71
+                - 83
+                - 95
+                - 98
+                - 100
+                - 105
+            "2":
+                - 1
+                - 2
+                - 3
+                - 5
+                - 11
+                - 12
+                - 41
+                - 49
+                - 50
+                - 51
+                - 53
+                - 55
+                - 71
+                - 83
+                - 95
+                - 98
+                - 100
+                - 105
+            "3":
+                - 13
+                - 15
+                - 16
+                - 23
+                - 55
+                - 61
+            "4": 3.9.25
+            "5": 0.18.6
+            "6": 4.47.1
+            "8":
+                - 5
+            "12": 0.18.6
+            "13": linux-x86_64
+action_dim:
+    value: 2
+checkpoint:
+    value:
+        save_last_ckpt: true
+        save_last_snapshot: false
+        topk:
+            format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
+            k: 5
+            mode: max
+            monitor_key: test_success_rate
+dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: true
+ema:
+    value:
+        _target_: llmbc.model.diffusion.ema_model.EMAModel
+        inv_gamma: 1
+        max_value: 0.9999
+        min_value: 0
+        power: 0.75
+        update_after_step: 0
+exp_name:
+    value: default
+horizon:
+    value: 1
+keypoint_visible_rate:
+    value: 1
+llm:
+    value:
+        causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+        checkpoint: data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850
+        config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+        finetune_mode: orig
+        hydra:
+            job:
+                override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+            run:
+                dir: data/outputs/2025.12.20/17.23.18_HuggingFaceTB/SmolLM2-135M-Instruct
+        llm_mode: ete-finetuned
+        lora_config:
+            bias: none
+            lora_alpha: 64
+            lora_dropout: 0.05
+            r: 32
+            task_type: CAUSAL_LM
+        max_length: 100
+        model_name: SmolLM2-135M-Instruct
+        name: HuggingFaceTB/SmolLM2-135M-Instruct
+        prompter:
+            _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+            use_joint_mlp_projector: true
+        use_joint_mlp_projector: true
+        use_quantization: false
+llm_do_sample:
+    value: false
+llm_orig_expert_feedback:
+    value: true
+logging:
+    value:
+        group: null
+        id: null
+        mode: online
+        name: 2025.12.20-17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+        project: maze2d-medium-v0-training
+        resume: true
+        tags:
+            - train_llm_diffusion_unet_lowdim
+            - maze2d-medium-v0
+            - default
+model_name:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+multi_run:
+    value:
+        run_dir: data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+        wandb_name_base: 2025.12.20-17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+n_action_steps:
+    value: 1
+n_latency_steps:
+    value: 0
+n_obs_steps:
+    value: 1
+name:
+    value: train_llm_diffusion_unet_lowdim
+obs_as_global_cond:
+    value: true
+obs_as_local_cond:
+    value: false
+obs_dim:
+    value: 6
+optimizer:
+    value:
+        _target_: torch.optim.AdamW
+        betas:
+            - 0.95
+            - 0.999
+        eps: 1e-08
+        lr: 0.0001
+        weight_decay: 1e-06
+output_dir:
+    value: /workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+past_action_visible:
+    value: false
+policy:
+    value:
+        _target_: llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy
+        action_dim: 2
+        horizon: 1
+        llm_discriminator:
+            _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+            llm_translator:
+                _target_: llmbc.translator.llm_translator.LLMTranslator
+                action_dim: 2
+                cfg:
+                    causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+                    checkpoint: data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850
+                    config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+                    finetune_mode: orig
+                    hydra:
+                        job:
+                            override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+                        run:
+                            dir: data/outputs/2025.12.20/17.23.18_HuggingFaceTB/SmolLM2-135M-Instruct
+                    llm_mode: ete-finetuned
+                    lora_config:
+                        bias: none
+                        lora_alpha: 64
+                        lora_dropout: 0.05
+                        r: 32
+                        task_type: CAUSAL_LM
+                    max_length: 100
+                    model_name: SmolLM2-135M-Instruct
+                    name: HuggingFaceTB/SmolLM2-135M-Instruct
+                    prompter:
+                        _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+                        use_joint_mlp_projector: true
+                    use_joint_mlp_projector: true
+                    use_quantization: false
+                horizon: 1
+                n_action_steps: 1
+                n_obs_steps: 1
+                obs_dim: 6
+            task_id: maze2d-medium-v0
+        loss_dp_weight: 1
+        loss_llm_weight: 0.0001
+        model:
+            _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
+            cond_predict_scale: true
+            diffusion_step_embed_dim: 256
+            down_dims:
+                - 256
+                - 512
+                - 1024
+            global_cond_dim: 6
+            input_dim: 2
+            kernel_size: 5
+            local_cond_dim: null
+            n_groups: 8
+        n_action_steps: 1
+        n_obs_steps: 1
+        noise_scheduler:
+            _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
+            beta_end: 0.02
+            beta_schedule: squaredcos_cap_v2
+            beta_start: 0.0001
+            clip_sample: true
+            num_train_timesteps: 100
+            prediction_type: epsilon
+            variance_type: fixed_small
+        normalize_llm_loss: true
+        num_inference_steps: 100
+        oa_step_convention: true
+        obs_as_global_cond: true
+        obs_as_local_cond: false
+        obs_dim: 6
+        pred_action_steps_only: false
+        reweight_llm_loss: true
+pred_action_steps_only:
+    value: false
+task:
+    value:
+        action_dim: 2
+        dataset:
+            _target_: llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset
+            data_path: datasets/maze2d-medium-v0.pt
+            data_path2: datasets/maze2d-medium-v0.pt
+            dummy_normalizer: false
+            horizon: 1
+            obs_eef_target: true
+            pad_after: 0
+            pad_before: 0
+            use_manual_normalizer: false
+            val_ratio: 0.02
+        env_runner:
+            _target_: llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner
+            discount: 0.99
+            env_name: llf-pointmaze-maze2d-medium-v0
+            feedback_type:
+                - hp
+                - hn
+                - fp
+            instruction_type: b
+            max_steps: 250
+            n_action_steps: 1
+            n_envs: 10
+            n_obs_steps: 1
+            n_test: 50
+            n_train: 10
+            visual: false
+        instructor:
+            _target_: llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor
+        name: maze2d-medium-v0
+        obs_dim: 6
+task_name:
+    value: maze2d-medium-v0
+training:
+    value:
+        checkpoint_every: 5
+        debug: false
+        device: cuda:0
+        gradient_accumulate_every: 16
+        lr_scheduler: cosine
+        lr_warmup_steps: 500
+        max_train_steps: null
+        max_val_steps: null
+        num_epochs: 1001
+        resume: true
+        rollout_every: 5
+        sample_every: 5
+        seed: 43
+        tqdm_interval_sec: 1
+        use_ema: true
+        val_every: 1
+val_dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: false

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/files/output.log

@@ -0,0 +1,5 @@
+/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/gym/envs/registration.py:505: UserWarning: WARN: The environment maze2d-medium-v0 is out of date. You should consider upgrading to version `v1` with the environment ID `maze2d-medium-v1`.
+  logger.warn(
+Eval PointmazeLowdimRunner 1/6:   0%|                                                       | 0/250 [00:00<?, ?it/s]/workspace/LLM-BC/llmbc/common/llfbench_util.py:39: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:275.)
+  obs = torch.tensor(obs, dtype=torch.float32).unsqueeze(dim=0).to(device)
+                                                                                                                    

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/files/wandb-metadata.json

@@ -0,0 +1,57 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.31",
+  "python": "3.9.25",
+  "startedAt": "2025-12-20T17:23:40.390717Z",
+  "args": [
+    "--config-path",
+    "./config/main_table",
+    "--config-name",
+    "llmdp_maze2d-medium-v0.yaml",
+    "policy.loss_llm_weight=1.0e-4",
+    "training.seed=43"
+  ],
+  "program": "/workspace/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "e3e08b12f4d28b96fced8dd2dc01206f0994f329"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0",
+  "host": "llmbc-cuda-2-6b87fc9b49-jfjww",
+  "username": "root",
+  "executable": "/root/miniconda3/envs/llm-bc/bin/python3",
+  "codePathLocal": "train.py",
+  "cpu_count": 64,
+  "cpu_count_logical": 128,
+  "gpu": "NVIDIA H100 80GB HBM3",
+  "gpu_count": 2,
+  "disk": {
+    "/": {
+      "total": "1918181408768",
+      "used": "1504054423552"
+    }
+  },
+  "memory": {
+    "total": "2163619762176"
+  },
+  "cpu": {
+    "count": 64,
+    "countLogical": 128
+  },
+  "gpu_nvidia": [
+    {
+      "name": "NVIDIA H100 80GB HBM3",
+      "memoryTotal": "85520809984",
+      "cudaCores": 16896,
+      "architecture": "Hopper"
+    },
+    {
+      "name": "NVIDIA H100 80GB HBM3",
+      "memoryTotal": "85520809984",
+      "cudaCores": 16896,
+      "architecture": "Hopper"
+    }
+  ],
+  "cudaVersion": "12.8"
+}

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/files/wandb-summary.json

@@ -0,0 +1 @@
+{"_wandb":{"runtime":614783},"lr":2.5315954887972227e-14,"_runtime":614783.19045473,"test/cumulative_reward":3.786958879193524,"epoch":1000,"test/mean_score":0.96,"val_loss":0.455483615398407,"_timestamp":1.766866201590722e+09,"test/success_rate":0.92,"train_loss_llm":0.0009613037109375,"_step":1587585,"train_action_mse_error":4.75653178000357e-05,"train_loss":0.005340841047547167,"global_step":1587585,"train/success_rate":0.9,"train/mean_score":0.9,"train/cumulative_reward":3.856401466699334,"train_loss_dp":0.0005138660199008882}

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug-core.log

@@ -0,0 +1,14 @@
+{"time":"2025-12-20T17:23:39.799221942Z","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpgr4g4w43/port-14556.txt","pid":14556,"debug":false,"disable-analytics":false}
+{"time":"2025-12-20T17:23:39.799250435Z","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2025-12-20T17:23:39.799701585Z","level":"INFO","msg":"Will exit if parent process dies.","ppid":14556}
+{"time":"2025-12-20T17:23:39.799701261Z","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":41909,"Zone":""}}
+{"time":"2025-12-20T17:23:39.99378811Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:50668"}
+{"time":"2025-12-20T17:23:40.392348342Z","level":"INFO","msg":"handleInformInit: received","streamId":"t2ymaznj","id":"127.0.0.1:50668"}
+{"time":"2025-12-20T17:23:40.500113061Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"t2ymaznj","id":"127.0.0.1:50668"}
+{"time":"2025-12-27T20:10:03.581036654Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:50668"}
+{"time":"2025-12-27T20:10:03.581122589Z","level":"INFO","msg":"server is shutting down"}
+{"time":"2025-12-27T20:10:03.581116025Z","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:50668"}
+{"time":"2025-12-27T20:10:03.581184842Z","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:50668"}
+{"time":"2025-12-27T20:10:05.183578531Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:50668"}
+{"time":"2025-12-27T20:10:05.183602036Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:50668"}
+{"time":"2025-12-27T20:10:05.183612102Z","level":"INFO","msg":"server is closed"}

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug-internal.log

@@ -0,0 +1,28 @@
+{"time":"2025-12-20T17:23:40.39265505Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-20T17:23:40.392690291Z","level":"INFO","msg":"created symlink","path":"/workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug-core.log"}
+{"time":"2025-12-20T17:23:40.50006116Z","level":"INFO","msg":"created new stream","id":"t2ymaznj"}
+{"time":"2025-12-20T17:23:40.500106816Z","level":"INFO","msg":"stream: started","id":"t2ymaznj"}
+{"time":"2025-12-20T17:23:40.500367774Z","level":"INFO","msg":"sender: started","stream_id":"t2ymaznj"}
+{"time":"2025-12-20T17:23:40.500120299Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-20T17:23:40.500143338Z","level":"INFO","msg":"handler: started","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-20T17:23:41.157625473Z","level":"INFO","msg":"Starting system monitor"}
+{"time":"2025-12-22T02:02:44.208411298Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:02:59.788100106Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": read tcp 192.168.19.170:48608->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-22T05:06:59.165164769Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T05:09:29.278598582Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T06:16:10.957017478Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T06:17:26.36373148Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:24:33.296938897Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:24:44.346922515Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:16.775885945Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:51.300218869Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:28:59.360396234Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:32:46.734796863Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/t2ymaznj/file_stream\": read tcp 192.168.19.170:44740->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-27T20:10:03.581119512Z","level":"INFO","msg":"stream: closing","id":"t2ymaznj"}
+{"time":"2025-12-27T20:10:03.581183738Z","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2025-12-27T20:10:03.583096625Z","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2025-12-27T20:10:04.923009029Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2025-12-27T20:10:05.183243732Z","level":"INFO","msg":"handler: closed","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-27T20:10:05.183318468Z","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"t2ymaznj"}}
+{"time":"2025-12-27T20:10:05.183326493Z","level":"INFO","msg":"sender: closed","stream_id":"t2ymaznj"}
+{"time":"2025-12-27T20:10:05.18352636Z","level":"INFO","msg":"stream: closed","id":"t2ymaznj"}

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug.log

@@ -0,0 +1,27 @@
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Configure stats pid to 14556
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Loading settings from /workspace/LLM-BC/wandb/settings
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/workspace/LLM-BC/train.py', 'program': '/workspace/LLM-BC/./train.py'}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:_log_setup():533] Logging user logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug.log
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:_log_setup():534] Logging internal logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/logs/debug-internal.log
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:init():619] calling init triggers
+2025-12-20 17:23:40,388 INFO    MainThread:14556 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llm_diffusion_unet_lowdim', '_target_': 'llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace', 'obs_dim': 6, 'action_dim': 2, 'task_name': 'maze2d-medium-v0', '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, 'keypoint_visible_rate': 1.0, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 2, 'local_cond_dim': None, 'global_cond_dim': 6, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'horizon': 1, 'obs_dim': 6, 'action_dim': 2, 'n_action_steps': 1, 'n_obs_steps': 1, 'num_inference_steps': 100, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'oa_step_convention': True, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'maze2d-medium-v0', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.18_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 6, 'action_dim': 2, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.0001, 'normalize_llm_loss': True, 'reweight_llm_loss': True}, 'ema': {'_target_': 'llmbc.model.diffusion.ema_model.EMAModel', 'update_after_step': 0, 'inv_gamma': 1.0, 'power': 0.75, 'min_value': 0.0, 'max_value': 0.9999}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': False, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 43, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'num_epochs': 1001, 'gradient_accumulate_every': 16, 'use_ema': True, 'rollout_every': 5, 'checkpoint_every': 5, 'val_every': 1, 'sample_every': 5, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'maze2d-medium-v0-training', 'resume': True, 'mode': 'online', 'name': '2025.12.20-17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'tags': ['train_llm_diffusion_unet_lowdim', 'maze2d-medium-v0', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'wandb_name_base': '2025.12.20-17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}, 'task': {'name': 'maze2d-medium-v0', 'obs_dim': 6, 'action_dim': 2, 'env_runner': {'_target_': 'llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner', 'env_name': 'llf-pointmaze-maze2d-medium-v0', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 250, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset', 'data_path': 'datasets/maze2d-medium-v0.pt', 'data_path2': 'datasets/maze2d-medium-v0.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.18_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-20 17:23:40,389 INFO    MainThread:14556 [wandb_init.py:init():669] starting backend
+2025-12-20 17:23:40,389 INFO    MainThread:14556 [wandb_init.py:init():673] sending inform_init request
+2025-12-20 17:23:40,390 INFO    MainThread:14556 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-20 17:23:40,390 INFO    MainThread:14556 [wandb_init.py:init():686] backend started and connected
+2025-12-20 17:23:40,397 INFO    MainThread:14556 [wandb_init.py:init():781] updated telemetry
+2025-12-20 17:23:40,470 INFO    MainThread:14556 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-20 17:23:41,128 INFO    MainThread:14556 [wandb_init.py:init():867] starting run threads in backend
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-20 17:23:42,555 INFO    MainThread:14556 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-20 17:23:42,624 INFO    MainThread:14556 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-20 17:23:42,624 INFO    MainThread:14556 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/workspace/LLM-BC/data/outputs/2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}
+2025-12-27 20:10:03,581 WARNING MsgRouterThr:14556 [router.py:message_loop():75] message_loop has been closed

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172340-t2ymaznj/run-t2ymaznj.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f63efb4d3576e44f1272052cfd2a0a511826ee5cbcf1a09511caa6ae4e69d3a
+size 1555253254

2025.12.20/17.23.18_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/wandb-resume.json

@@ -0,0 +1 @@
+{"run_id": "t2ymaznj"}

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/config.yaml

@@ -0,0 +1,193 @@
+name: train_llm_diffusion_unet_lowdim
+_target_: llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace
+obs_dim: ${task.obs_dim}
+action_dim: ${task.action_dim}
+task_name: ${task.name}
+exp_name: default
+model_name: ${llm.name}
+horizon: 1
+n_obs_steps: 1
+n_action_steps: 1
+n_latency_steps: 0
+past_action_visible: false
+keypoint_visible_rate: 1.0
+obs_as_local_cond: false
+obs_as_global_cond: true
+pred_action_steps_only: false
+llm_orig_expert_feedback: true
+llm_do_sample: false
+policy:
+  _target_: llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy
+  model:
+    _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
+    input_dim: '${eval: ${task.action_dim} if ${obs_as_local_cond} or ${obs_as_global_cond}
+      else ${task.obs_dim} + ${task.action_dim}}'
+    local_cond_dim: '${eval: ${task.obs_dim} if ${obs_as_local_cond} else None}'
+    global_cond_dim: '${eval: ${task.obs_dim}*${n_obs_steps} if ${obs_as_global_cond}
+      else None}'
+    diffusion_step_embed_dim: 256
+    down_dims:
+    - 256
+    - 512
+    - 1024
+    kernel_size: 5
+    n_groups: 8
+    cond_predict_scale: true
+  noise_scheduler:
+    _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
+    num_train_timesteps: 100
+    beta_start: 0.0001
+    beta_end: 0.02
+    beta_schedule: squaredcos_cap_v2
+    variance_type: fixed_small
+    clip_sample: true
+    prediction_type: epsilon
+  horizon: ${horizon}
+  obs_dim: ${obs_dim}
+  action_dim: ${action_dim}
+  n_action_steps: ${eval:'${n_action_steps}+${n_latency_steps}'}
+  n_obs_steps: ${n_obs_steps}
+  num_inference_steps: 100
+  obs_as_local_cond: ${obs_as_local_cond}
+  obs_as_global_cond: ${obs_as_global_cond}
+  pred_action_steps_only: ${pred_action_steps_only}
+  oa_step_convention: true
+  llm_discriminator:
+    _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+    task_id: ${task_name}
+    llm_translator:
+      _target_: llmbc.translator.llm_translator.LLMTranslator
+      cfg: ${llm}
+      obs_dim: ${obs_dim}
+      action_dim: ${action_dim}
+      horizon: ${horizon}
+      n_obs_steps: ${n_obs_steps}
+      n_action_steps: ${eval:'${n_action_steps}+${n_latency_steps}'}
+  loss_dp_weight: 1.0
+  loss_llm_weight: 0.0001
+  normalize_llm_loss: true
+  reweight_llm_loss: true
+ema:
+  _target_: llmbc.model.diffusion.ema_model.EMAModel
+  update_after_step: 0
+  inv_gamma: 1.0
+  power: 0.75
+  min_value: 0.0
+  max_value: 0.9999
+dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: false
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.AdamW
+  lr: 0.0001
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda:0
+  seed: 44
+  debug: false
+  resume: true
+  lr_scheduler: cosine
+  lr_warmup_steps: 500
+  num_epochs: 1001
+  gradient_accumulate_every: 16
+  use_ema: true
+  rollout_every: 5
+  checkpoint_every: 5
+  val_every: 1
+  sample_every: 5
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: ${task.name}-training
+  resume: true
+  mode: online
+  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+  tags:
+  - ${name}
+  - ${task_name}
+  - ${exp_name}
+  id: null
+  group: null
+checkpoint:
+  topk:
+    monitor_key: test_success_rate
+    mode: max
+    k: 5
+    format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
+  save_last_ckpt: true
+  save_last_snapshot: false
+multi_run:
+  run_dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  wandb_name_base: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+task:
+  name: maze2d-medium-v0
+  obs_dim: 6
+  action_dim: 2
+  env_runner:
+    _target_: llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner
+    env_name: llf-pointmaze-maze2d-medium-v0
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 250
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    - fp
+    visual: false
+    discount: 0.99
+  dataset:
+    _target_: llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset
+    data_path: datasets/maze2d-medium-v0.pt
+    data_path2: datasets/maze2d-medium-v0.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: 0.02
+    dummy_normalizer: false
+  instructor:
+    _target_: llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor
+llm:
+  name: HuggingFaceTB/SmolLM2-135M-Instruct
+  model_name: SmolLM2-135M-Instruct
+  config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+  causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+  use_quantization: false
+  use_joint_mlp_projector: true
+  llm_mode: ete-finetuned
+  finetune_mode: orig
+  checkpoint: data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850
+  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}

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/hydra.yaml

@@ -0,0 +1,156 @@
+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:
+    - policy.loss_llm_weight=1.0e-4
+    - training.seed=44
+  job:
+    name: train
+    chdir: null
+    override_dirname: policy.loss_llm_weight=1.0e-4,training.seed=44
+    id: ???
+    num: ???
+    config_name: llmdp_maze2d-medium-v0.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+    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

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/overrides.yaml

@@ -0,0 +1,2 @@
+- policy.loss_llm_weight=1.0e-4
+- training.seed=44

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0385-test_success_rate=0.900.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6faf3353a17defbc850ea4fca4b452fedc4c9706940c2d9b03b48ea46f873ecf
+size 1039753586

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0485-test_success_rate=0.900.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ef7d28f09e0517e0a646a5de7c7de90e444617a849639f12c6e6fc1d9d312b2
+size 1039753586

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0635-test_success_rate=0.900.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6dea66019eafff0edd9d1abe866682c76022efdfe66b671f3ba05b75c97b824
+size 1039753586

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0670-test_success_rate=0.900.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3c46bb3e1707cff883510aa1150dadd5ddf52a86f4912df0cdd41ba71940138
+size 1039753586

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0680-test_success_rate=0.920.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba5ca8daeab5bdac5c908bc621d9c353fb06a937179384802168314c6e6f9be0
+size 1039753586

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/latest.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06113f2c593ffde672569d6f7b68a81d411e89b3d9abf1fb92289fd0419bda55
+size 1039753586

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/logs.json.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf8bebd9c8a93c2df953f1617225eb423cb84a8a4bf136d3db48da96981b5ebe
+size 279217611

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/train.log

@@ -0,0 +1,9 @@
+[2025-12-20 17:23:24,946][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2025-12-20 17:23:24,947][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2025-12-20 17:23:24,947][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2025-12-20 17:23:30,016][datasets][INFO] - PyTorch version 2.2.2 available.
+[2025-12-20 17:23:30,017][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2025-12-20 17:23:30,018][datasets][INFO] - JAX version 0.4.30 available.
+[2025-12-20 17:23:31,691][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.496819e+07
+[2025-12-20 17:23:45,647][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
+[2025-12-20 17:23:45,657][absl][INFO] - MuJoCo library version is: 2.3.7

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/debug-internal.log

@@ -0,0 +1,35 @@
+{"time":"2025-12-20T17:23:46.613199748Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-20T17:23:46.613264827Z","level":"INFO","msg":"created symlink","path":"/workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug-core.log"}
+{"time":"2025-12-20T17:23:46.72929786Z","level":"INFO","msg":"created new stream","id":"cflo0geg"}
+{"time":"2025-12-20T17:23:46.729364477Z","level":"INFO","msg":"stream: started","id":"cflo0geg"}
+{"time":"2025-12-20T17:23:46.729395032Z","level":"INFO","msg":"sender: started","stream_id":"cflo0geg"}
+{"time":"2025-12-20T17:23:46.72938452Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-20T17:23:46.729401223Z","level":"INFO","msg":"handler: started","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-20T17:23:47.379019149Z","level":"INFO","msg":"Starting system monitor"}
+{"time":"2025-12-22T01:58:59.359063808Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T02:00:48.535421007Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:01:19.054517009Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T02:01:20.671973451Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:01:55.142789828Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:02:34.824004591Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:02:58.288135548Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": read tcp 192.168.19.170:33170->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-22T05:10:18.616078123Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T05:10:19.705729594Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": read tcp 192.168.19.170:40764->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-22T07:24:48.668983055Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:05.56897683Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:25:20.978463529Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:55.205939946Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:28:03.680647103Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:28:37.646706307Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": EOF"}
+{"time":"2025-12-22T07:29:16.409762506Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:32:33.686447472Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:33:05.887016615Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:35:49.545876342Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-27T19:54:59.326572801Z","level":"INFO","msg":"stream: closing","id":"cflo0geg"}
+{"time":"2025-12-27T19:54:59.326610662Z","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2025-12-27T19:54:59.340210928Z","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2025-12-27T19:55:02.304599868Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2025-12-27T19:55:02.563003931Z","level":"INFO","msg":"handler: closed","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-27T19:55:02.563038962Z","level":"INFO","msg":"sender: closed","stream_id":"cflo0geg"}
+{"time":"2025-12-27T19:55:02.563034816Z","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-27T19:55:02.563274828Z","level":"INFO","msg":"stream: closed","id":"cflo0geg"}

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/debug.log

@@ -0,0 +1,27 @@
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Configure stats pid to 14641
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Loading settings from /workspace/LLM-BC/wandb/settings
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/workspace/LLM-BC/train.py', 'program': '/workspace/LLM-BC/./train.py'}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:_log_setup():533] Logging user logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug.log
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:_log_setup():534] Logging internal logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug-internal.log
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():619] calling init triggers
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llm_diffusion_unet_lowdim', '_target_': 'llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace', 'obs_dim': 6, 'action_dim': 2, 'task_name': 'maze2d-medium-v0', '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, 'keypoint_visible_rate': 1.0, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 2, 'local_cond_dim': None, 'global_cond_dim': 6, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'horizon': 1, 'obs_dim': 6, 'action_dim': 2, 'n_action_steps': 1, 'n_obs_steps': 1, 'num_inference_steps': 100, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'oa_step_convention': True, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'maze2d-medium-v0', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.23_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 6, 'action_dim': 2, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.0001, 'normalize_llm_loss': True, 'reweight_llm_loss': True}, 'ema': {'_target_': 'llmbc.model.diffusion.ema_model.EMAModel', 'update_after_step': 0, 'inv_gamma': 1.0, 'power': 0.75, 'min_value': 0.0, 'max_value': 0.9999}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': False, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 44, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'num_epochs': 1001, 'gradient_accumulate_every': 16, 'use_ema': True, 'rollout_every': 5, 'checkpoint_every': 5, 'val_every': 1, 'sample_every': 5, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'maze2d-medium-v0-training', 'resume': True, 'mode': 'online', 'name': '2025.12.20-17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'tags': ['train_llm_diffusion_unet_lowdim', 'maze2d-medium-v0', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'wandb_name_base': '2025.12.20-17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}, 'task': {'name': 'maze2d-medium-v0', 'obs_dim': 6, 'action_dim': 2, 'env_runner': {'_target_': 'llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner', 'env_name': 'llf-pointmaze-maze2d-medium-v0', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 250, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset', 'data_path': 'datasets/maze2d-medium-v0.pt', 'data_path2': 'datasets/maze2d-medium-v0.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.23_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():669] starting backend
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():673] sending inform_init request
+2025-12-20 17:23:46,610 INFO    MainThread:14641 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-20 17:23:46,610 INFO    MainThread:14641 [wandb_init.py:init():686] backend started and connected
+2025-12-20 17:23:46,618 INFO    MainThread:14641 [wandb_init.py:init():781] updated telemetry
+2025-12-20 17:23:46,663 INFO    MainThread:14641 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-20 17:23:47,376 INFO    MainThread:14641 [wandb_init.py:init():867] starting run threads in backend
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-20 17:23:47,752 INFO    MainThread:14641 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-20 17:23:47,753 INFO    MainThread:14641 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}
+2025-12-27 19:54:59,326 WARNING MsgRouterThr:14641 [router.py:message_loop():75] message_loop has been closed

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/files/config.yaml

@@ -0,0 +1,304 @@
+_target_:
+    value: llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace
+_wandb:
+    value:
+        cli_version: 0.18.6
+        m: []
+        python_version: 3.9.25
+        t:
+            "1":
+                - 1
+                - 2
+                - 3
+                - 5
+                - 11
+                - 12
+                - 41
+                - 49
+                - 50
+                - 51
+                - 53
+                - 55
+                - 71
+                - 83
+                - 95
+                - 98
+                - 100
+                - 105
+            "2":
+                - 1
+                - 2
+                - 3
+                - 5
+                - 11
+                - 12
+                - 41
+                - 49
+                - 50
+                - 51
+                - 53
+                - 55
+                - 71
+                - 83
+                - 95
+                - 98
+                - 100
+                - 105
+            "3":
+                - 13
+                - 15
+                - 16
+                - 23
+                - 55
+                - 61
+            "4": 3.9.25
+            "5": 0.18.6
+            "6": 4.47.1
+            "8":
+                - 5
+            "12": 0.18.6
+            "13": linux-x86_64
+action_dim:
+    value: 2
+checkpoint:
+    value:
+        save_last_ckpt: true
+        save_last_snapshot: false
+        topk:
+            format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
+            k: 5
+            mode: max
+            monitor_key: test_success_rate
+dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: true
+ema:
+    value:
+        _target_: llmbc.model.diffusion.ema_model.EMAModel
+        inv_gamma: 1
+        max_value: 0.9999
+        min_value: 0
+        power: 0.75
+        update_after_step: 0
+exp_name:
+    value: default
+horizon:
+    value: 1
+keypoint_visible_rate:
+    value: 1
+llm:
+    value:
+        causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+        checkpoint: data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850
+        config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+        finetune_mode: orig
+        hydra:
+            job:
+                override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+            run:
+                dir: data/outputs/2025.12.20/17.23.23_HuggingFaceTB/SmolLM2-135M-Instruct
+        llm_mode: ete-finetuned
+        lora_config:
+            bias: none
+            lora_alpha: 64
+            lora_dropout: 0.05
+            r: 32
+            task_type: CAUSAL_LM
+        max_length: 100
+        model_name: SmolLM2-135M-Instruct
+        name: HuggingFaceTB/SmolLM2-135M-Instruct
+        prompter:
+            _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+            use_joint_mlp_projector: true
+        use_joint_mlp_projector: true
+        use_quantization: false
+llm_do_sample:
+    value: false
+llm_orig_expert_feedback:
+    value: true
+logging:
+    value:
+        group: null
+        id: null
+        mode: online
+        name: 2025.12.20-17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+        project: maze2d-medium-v0-training
+        resume: true
+        tags:
+            - train_llm_diffusion_unet_lowdim
+            - maze2d-medium-v0
+            - default
+model_name:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+multi_run:
+    value:
+        run_dir: data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+        wandb_name_base: 2025.12.20-17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+n_action_steps:
+    value: 1
+n_latency_steps:
+    value: 0
+n_obs_steps:
+    value: 1
+name:
+    value: train_llm_diffusion_unet_lowdim
+obs_as_global_cond:
+    value: true
+obs_as_local_cond:
+    value: false
+obs_dim:
+    value: 6
+optimizer:
+    value:
+        _target_: torch.optim.AdamW
+        betas:
+            - 0.95
+            - 0.999
+        eps: 1e-08
+        lr: 0.0001
+        weight_decay: 1e-06
+output_dir:
+    value: /workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+past_action_visible:
+    value: false
+policy:
+    value:
+        _target_: llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy
+        action_dim: 2
+        horizon: 1
+        llm_discriminator:
+            _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+            llm_translator:
+                _target_: llmbc.translator.llm_translator.LLMTranslator
+                action_dim: 2
+                cfg:
+                    causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+                    checkpoint: data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850
+                    config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+                    finetune_mode: orig
+                    hydra:
+                        job:
+                            override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+                        run:
+                            dir: data/outputs/2025.12.20/17.23.23_HuggingFaceTB/SmolLM2-135M-Instruct
+                    llm_mode: ete-finetuned
+                    lora_config:
+                        bias: none
+                        lora_alpha: 64
+                        lora_dropout: 0.05
+                        r: 32
+                        task_type: CAUSAL_LM
+                    max_length: 100
+                    model_name: SmolLM2-135M-Instruct
+                    name: HuggingFaceTB/SmolLM2-135M-Instruct
+                    prompter:
+                        _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+                        use_joint_mlp_projector: true
+                    use_joint_mlp_projector: true
+                    use_quantization: false
+                horizon: 1
+                n_action_steps: 1
+                n_obs_steps: 1
+                obs_dim: 6
+            task_id: maze2d-medium-v0
+        loss_dp_weight: 1
+        loss_llm_weight: 0.0001
+        model:
+            _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
+            cond_predict_scale: true
+            diffusion_step_embed_dim: 256
+            down_dims:
+                - 256
+                - 512
+                - 1024
+            global_cond_dim: 6
+            input_dim: 2
+            kernel_size: 5
+            local_cond_dim: null
+            n_groups: 8
+        n_action_steps: 1
+        n_obs_steps: 1
+        noise_scheduler:
+            _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
+            beta_end: 0.02
+            beta_schedule: squaredcos_cap_v2
+            beta_start: 0.0001
+            clip_sample: true
+            num_train_timesteps: 100
+            prediction_type: epsilon
+            variance_type: fixed_small
+        normalize_llm_loss: true
+        num_inference_steps: 100
+        oa_step_convention: true
+        obs_as_global_cond: true
+        obs_as_local_cond: false
+        obs_dim: 6
+        pred_action_steps_only: false
+        reweight_llm_loss: true
+pred_action_steps_only:
+    value: false
+task:
+    value:
+        action_dim: 2
+        dataset:
+            _target_: llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset
+            data_path: datasets/maze2d-medium-v0.pt
+            data_path2: datasets/maze2d-medium-v0.pt
+            dummy_normalizer: false
+            horizon: 1
+            obs_eef_target: true
+            pad_after: 0
+            pad_before: 0
+            use_manual_normalizer: false
+            val_ratio: 0.02
+        env_runner:
+            _target_: llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner
+            discount: 0.99
+            env_name: llf-pointmaze-maze2d-medium-v0
+            feedback_type:
+                - hp
+                - hn
+                - fp
+            instruction_type: b
+            max_steps: 250
+            n_action_steps: 1
+            n_envs: 10
+            n_obs_steps: 1
+            n_test: 50
+            n_train: 10
+            visual: false
+        instructor:
+            _target_: llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor
+        name: maze2d-medium-v0
+        obs_dim: 6
+task_name:
+    value: maze2d-medium-v0
+training:
+    value:
+        checkpoint_every: 5
+        debug: false
+        device: cuda:0
+        gradient_accumulate_every: 16
+        lr_scheduler: cosine
+        lr_warmup_steps: 500
+        max_train_steps: null
+        max_val_steps: null
+        num_epochs: 1001
+        resume: true
+        rollout_every: 5
+        sample_every: 5
+        seed: 44
+        tqdm_interval_sec: 1
+        use_ema: true
+        val_every: 1
+val_dataloader:
+    value:
+        batch_size: 16
+        num_workers: 0
+        persistent_workers: false
+        pin_memory: false
+        shuffle: false

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/files/output.log

@@ -0,0 +1,5 @@
+/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/gym/envs/registration.py:505: UserWarning: WARN: The environment maze2d-medium-v0 is out of date. You should consider upgrading to version `v1` with the environment ID `maze2d-medium-v1`.
+  logger.warn(
+Eval PointmazeLowdimRunner 1/6:   0%|                                                       | 0/250 [00:00<?, ?it/s]/workspace/LLM-BC/llmbc/common/llfbench_util.py:39: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:275.)
+  obs = torch.tensor(obs, dtype=torch.float32).unsqueeze(dim=0).to(device)
+                                                                                                                    

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/files/wandb-metadata.json

@@ -0,0 +1,57 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.31",
+  "python": "3.9.25",
+  "startedAt": "2025-12-20T17:23:46.611044Z",
+  "args": [
+    "--config-path",
+    "./config/main_table",
+    "--config-name",
+    "llmdp_maze2d-medium-v0.yaml",
+    "policy.loss_llm_weight=1.0e-4",
+    "training.seed=44"
+  ],
+  "program": "/workspace/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "e3e08b12f4d28b96fced8dd2dc01206f0994f329"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0",
+  "host": "llmbc-cuda-2-6b87fc9b49-jfjww",
+  "username": "root",
+  "executable": "/root/miniconda3/envs/llm-bc/bin/python3",
+  "codePathLocal": "train.py",
+  "cpu_count": 64,
+  "cpu_count_logical": 128,
+  "gpu": "NVIDIA H100 80GB HBM3",
+  "gpu_count": 2,
+  "disk": {
+    "/": {
+      "total": "1918181408768",
+      "used": "1504054484992"
+    }
+  },
+  "memory": {
+    "total": "2163619762176"
+  },
+  "cpu": {
+    "count": 64,
+    "countLogical": 128
+  },
+  "gpu_nvidia": [
+    {
+      "name": "NVIDIA H100 80GB HBM3",
+      "memoryTotal": "85520809984",
+      "cudaCores": 16896,
+      "architecture": "Hopper"
+    },
+    {
+      "name": "NVIDIA H100 80GB HBM3",
+      "memoryTotal": "85520809984",
+      "cudaCores": 16896,
+      "architecture": "Hopper"
+    }
+  ],
+  "cudaVersion": "12.8"
+}

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/files/wandb-summary.json

@@ -0,0 +1 @@
+{"_wandb":{"runtime":613872},"train/success_rate":0.8,"epoch":1000,"train/mean_score":0.9,"lr":2.5251967183947954e-14,"test/mean_score":0.92,"train/cumulative_reward":4.296642909961845,"_timestamp":1.7668652974649105e+09,"test/success_rate":0.82,"global_step":1589587,"_step":1589587,"train_loss":0.005287302996962823,"val_loss":0.26226067543029785,"train_loss_llm":0,"train_loss_dp":0.009851432405412197,"_runtime":613872.71555456,"train_action_mse_error":9.117730951402336e-05,"test/cumulative_reward":3.934787119516622}

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug-core.log

@@ -0,0 +1,14 @@
+{"time":"2025-12-20T17:23:46.032804955Z","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmp_cxy2kjz/port-14641.txt","pid":14641,"debug":false,"disable-analytics":false}
+{"time":"2025-12-20T17:23:46.032842867Z","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2025-12-20T17:23:46.129614882Z","level":"INFO","msg":"Will exit if parent process dies.","ppid":14641}
+{"time":"2025-12-20T17:23:46.129619244Z","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":43043,"Zone":""}}
+{"time":"2025-12-20T17:23:46.224857458Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:54640"}
+{"time":"2025-12-20T17:23:46.612444569Z","level":"INFO","msg":"handleInformInit: received","streamId":"cflo0geg","id":"127.0.0.1:54640"}
+{"time":"2025-12-20T17:23:46.729372026Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"cflo0geg","id":"127.0.0.1:54640"}
+{"time":"2025-12-27T19:54:59.326504829Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:54640"}
+{"time":"2025-12-27T19:54:59.326587297Z","level":"INFO","msg":"server is shutting down"}
+{"time":"2025-12-27T19:54:59.326579185Z","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:54640"}
+{"time":"2025-12-27T19:54:59.326655838Z","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:54640"}
+{"time":"2025-12-27T19:55:02.563329274Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:54640"}
+{"time":"2025-12-27T19:55:02.563348986Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:54640"}
+{"time":"2025-12-27T19:55:02.563356697Z","level":"INFO","msg":"server is closed"}

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug-internal.log

@@ -0,0 +1,35 @@
+{"time":"2025-12-20T17:23:46.613199748Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-20T17:23:46.613264827Z","level":"INFO","msg":"created symlink","path":"/workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug-core.log"}
+{"time":"2025-12-20T17:23:46.72929786Z","level":"INFO","msg":"created new stream","id":"cflo0geg"}
+{"time":"2025-12-20T17:23:46.729364477Z","level":"INFO","msg":"stream: started","id":"cflo0geg"}
+{"time":"2025-12-20T17:23:46.729395032Z","level":"INFO","msg":"sender: started","stream_id":"cflo0geg"}
+{"time":"2025-12-20T17:23:46.72938452Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-20T17:23:46.729401223Z","level":"INFO","msg":"handler: started","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-20T17:23:47.379019149Z","level":"INFO","msg":"Starting system monitor"}
+{"time":"2025-12-22T01:58:59.359063808Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T02:00:48.535421007Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:01:19.054517009Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T02:01:20.671973451Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:01:55.142789828Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:02:34.824004591Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T02:02:58.288135548Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": read tcp 192.168.19.170:33170->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-22T05:10:18.616078123Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T05:10:19.705729594Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": read tcp 192.168.19.170:40764->35.186.228.49:443: read: connection reset by peer"}
+{"time":"2025-12-22T07:24:48.668983055Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:05.56897683Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:25:20.978463529Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:25:55.205939946Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:28:03.680647103Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:28:37.646706307Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": EOF"}
+{"time":"2025-12-22T07:29:16.409762506Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-22T07:32:33.686447472Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": net/http: request canceled (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:33:05.887016615Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/graphql\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"}
+{"time":"2025-12-22T07:35:49.545876342Z","level":"INFO","msg":"api: retrying error","error":"Post \"https://api.wandb.ai/files/chyang25-national-taiwan-university/maze2d-medium-v0-training/cflo0geg/file_stream\": unexpected EOF"}
+{"time":"2025-12-27T19:54:59.326572801Z","level":"INFO","msg":"stream: closing","id":"cflo0geg"}
+{"time":"2025-12-27T19:54:59.326610662Z","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2025-12-27T19:54:59.340210928Z","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2025-12-27T19:55:02.304599868Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2025-12-27T19:55:02.563003931Z","level":"INFO","msg":"handler: closed","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-27T19:55:02.563038962Z","level":"INFO","msg":"sender: closed","stream_id":"cflo0geg"}
+{"time":"2025-12-27T19:55:02.563034816Z","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"cflo0geg"}}
+{"time":"2025-12-27T19:55:02.563274828Z","level":"INFO","msg":"stream: closed","id":"cflo0geg"}

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug.log

@@ -0,0 +1,27 @@
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Configure stats pid to 14641
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Loading settings from /workspace/LLM-BC/wandb/settings
+2025-12-20 17:23:46,608 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/workspace/LLM-BC/train.py', 'program': '/workspace/LLM-BC/./train.py'}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:_log_setup():533] Logging user logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug.log
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:_log_setup():534] Logging internal logs to /workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/logs/debug-internal.log
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():619] calling init triggers
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llm_diffusion_unet_lowdim', '_target_': 'llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace', 'obs_dim': 6, 'action_dim': 2, 'task_name': 'maze2d-medium-v0', '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, 'keypoint_visible_rate': 1.0, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'llm_orig_expert_feedback': True, 'llm_do_sample': False, 'policy': {'_target_': 'llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy', 'model': {'_target_': 'llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D', 'input_dim': 2, 'local_cond_dim': None, 'global_cond_dim': 6, 'diffusion_step_embed_dim': 256, 'down_dims': [256, 512, 1024], 'kernel_size': 5, 'n_groups': 8, 'cond_predict_scale': True}, 'noise_scheduler': {'_target_': 'diffusers.schedulers.scheduling_ddpm.DDPMScheduler', 'num_train_timesteps': 100, 'beta_start': 0.0001, 'beta_end': 0.02, 'beta_schedule': 'squaredcos_cap_v2', 'variance_type': 'fixed_small', 'clip_sample': True, 'prediction_type': 'epsilon'}, 'horizon': 1, 'obs_dim': 6, 'action_dim': 2, 'n_action_steps': 1, 'n_obs_steps': 1, 'num_inference_steps': 100, 'obs_as_local_cond': False, 'obs_as_global_cond': True, 'pred_action_steps_only': False, 'oa_step_convention': True, 'llm_discriminator': {'_target_': 'llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator', 'task_id': 'maze2d-medium-v0', 'llm_translator': {'_target_': 'llmbc.translator.llm_translator.LLMTranslator', 'cfg': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.23_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 6, 'action_dim': 2, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.0001, 'normalize_llm_loss': True, 'reweight_llm_loss': True}, 'ema': {'_target_': 'llmbc.model.diffusion.ema_model.EMAModel', 'update_after_step': 0, 'inv_gamma': 1.0, 'power': 0.75, 'min_value': 0.0, 'max_value': 0.9999}, 'dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': True, 'pin_memory': False, 'persistent_workers': False}, 'val_dataloader': {'batch_size': 16, 'num_workers': 0, 'shuffle': False, 'pin_memory': False, 'persistent_workers': False}, 'optimizer': {'_target_': 'torch.optim.AdamW', 'lr': 0.0001, 'betas': [0.95, 0.999], 'eps': 1e-08, 'weight_decay': 1e-06}, 'training': {'device': 'cuda:0', 'seed': 44, 'debug': False, 'resume': True, 'lr_scheduler': 'cosine', 'lr_warmup_steps': 500, 'num_epochs': 1001, 'gradient_accumulate_every': 16, 'use_ema': True, 'rollout_every': 5, 'checkpoint_every': 5, 'val_every': 1, 'sample_every': 5, 'max_train_steps': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'maze2d-medium-v0-training', 'resume': True, 'mode': 'online', 'name': '2025.12.20-17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'tags': ['train_llm_diffusion_unet_lowdim', 'maze2d-medium-v0', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0', 'wandb_name_base': '2025.12.20-17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}, 'task': {'name': 'maze2d-medium-v0', 'obs_dim': 6, 'action_dim': 2, 'env_runner': {'_target_': 'llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner', 'env_name': 'llf-pointmaze-maze2d-medium-v0', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 250, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset', 'data_path': 'datasets/maze2d-medium-v0.pt', 'data_path2': 'datasets/maze2d-medium-v0.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'ete-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850', '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/2025.12.20/17.23.23_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():669] starting backend
+2025-12-20 17:23:46,609 INFO    MainThread:14641 [wandb_init.py:init():673] sending inform_init request
+2025-12-20 17:23:46,610 INFO    MainThread:14641 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-20 17:23:46,610 INFO    MainThread:14641 [wandb_init.py:init():686] backend started and connected
+2025-12-20 17:23:46,618 INFO    MainThread:14641 [wandb_init.py:init():781] updated telemetry
+2025-12-20 17:23:46,663 INFO    MainThread:14641 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-20 17:23:47,376 INFO    MainThread:14641 [wandb_init.py:init():867] starting run threads in backend
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-20 17:23:47,750 INFO    MainThread:14641 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-20 17:23:47,752 INFO    MainThread:14641 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-20 17:23:47,753 INFO    MainThread:14641 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/workspace/LLM-BC/data/outputs/2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0'}
+2025-12-27 19:54:59,326 WARNING MsgRouterThr:14641 [router.py:message_loop():75] message_loop has been closed

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/run-20251220_172346-cflo0geg/run-cflo0geg.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98ec254f012d18dfeb85ec2cbb996eb8d79f69a4b99fa6fccf958a6472c3995b
+size 1556242175

2025.12.20/17.23.23_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/wandb/wandb-resume.json

@@ -0,0 +1 @@
+{"run_id": "cflo0geg"}

2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/config.yaml

@@ -0,0 +1,193 @@
+name: train_llm_diffusion_unet_lowdim
+_target_: llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace
+obs_dim: ${task.obs_dim}
+action_dim: ${task.action_dim}
+task_name: ${task.name}
+exp_name: default
+model_name: ${llm.name}
+horizon: 1
+n_obs_steps: 1
+n_action_steps: 1
+n_latency_steps: 0
+past_action_visible: false
+keypoint_visible_rate: 1.0
+obs_as_local_cond: false
+obs_as_global_cond: true
+pred_action_steps_only: false
+llm_orig_expert_feedback: true
+llm_do_sample: false
+policy:
+  _target_: llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy
+  model:
+    _target_: llmbc.model.diffusion.conditional_unet1d.ConditionalUnet1D
+    input_dim: '${eval: ${task.action_dim} if ${obs_as_local_cond} or ${obs_as_global_cond}
+      else ${task.obs_dim} + ${task.action_dim}}'
+    local_cond_dim: '${eval: ${task.obs_dim} if ${obs_as_local_cond} else None}'
+    global_cond_dim: '${eval: ${task.obs_dim}*${n_obs_steps} if ${obs_as_global_cond}
+      else None}'
+    diffusion_step_embed_dim: 256
+    down_dims:
+    - 256
+    - 512
+    - 1024
+    kernel_size: 5
+    n_groups: 8
+    cond_predict_scale: true
+  noise_scheduler:
+    _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler
+    num_train_timesteps: 100
+    beta_start: 0.0001
+    beta_end: 0.02
+    beta_schedule: squaredcos_cap_v2
+    variance_type: fixed_small
+    clip_sample: true
+    prediction_type: epsilon
+  horizon: ${horizon}
+  obs_dim: ${obs_dim}
+  action_dim: ${action_dim}
+  n_action_steps: ${eval:'${n_action_steps}+${n_latency_steps}'}
+  n_obs_steps: ${n_obs_steps}
+  num_inference_steps: 100
+  obs_as_local_cond: ${obs_as_local_cond}
+  obs_as_global_cond: ${obs_as_global_cond}
+  pred_action_steps_only: ${pred_action_steps_only}
+  oa_step_convention: true
+  llm_discriminator:
+    _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+    task_id: ${task_name}
+    llm_translator:
+      _target_: llmbc.translator.llm_translator.LLMTranslator
+      cfg: ${llm}
+      obs_dim: ${obs_dim}
+      action_dim: ${action_dim}
+      horizon: ${horizon}
+      n_obs_steps: ${n_obs_steps}
+      n_action_steps: ${eval:'${n_action_steps}+${n_latency_steps}'}
+  loss_dp_weight: 1.0
+  loss_llm_weight: 0.0001
+  normalize_llm_loss: true
+  reweight_llm_loss: true
+ema:
+  _target_: llmbc.model.diffusion.ema_model.EMAModel
+  update_after_step: 0
+  inv_gamma: 1.0
+  power: 0.75
+  min_value: 0.0
+  max_value: 0.9999
+dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: true
+  pin_memory: false
+  persistent_workers: false
+val_dataloader:
+  batch_size: 16
+  num_workers: 0
+  shuffle: false
+  pin_memory: false
+  persistent_workers: false
+optimizer:
+  _target_: torch.optim.AdamW
+  lr: 0.0001
+  betas:
+  - 0.95
+  - 0.999
+  eps: 1.0e-08
+  weight_decay: 1.0e-06
+training:
+  device: cuda:0
+  seed: 45
+  debug: false
+  resume: true
+  lr_scheduler: cosine
+  lr_warmup_steps: 500
+  num_epochs: 1001
+  gradient_accumulate_every: 16
+  use_ema: true
+  rollout_every: 5
+  checkpoint_every: 5
+  val_every: 1
+  sample_every: 5
+  max_train_steps: null
+  max_val_steps: null
+  tqdm_interval_sec: 1.0
+logging:
+  project: ${task.name}-training
+  resume: true
+  mode: online
+  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+  tags:
+  - ${name}
+  - ${task_name}
+  - ${exp_name}
+  id: null
+  group: null
+checkpoint:
+  topk:
+    monitor_key: test_success_rate
+    mode: max
+    k: 5
+    format_str: epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt
+  save_last_ckpt: true
+  save_last_snapshot: false
+multi_run:
+  run_dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  wandb_name_base: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+task:
+  name: maze2d-medium-v0
+  obs_dim: 6
+  action_dim: 2
+  env_runner:
+    _target_: llmbc.env_runner.pointmaze_lowdim_runner.PointmazeLowdimRunner
+    env_name: llf-pointmaze-maze2d-medium-v0
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 250
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    - fp
+    visual: false
+    discount: 0.99
+  dataset:
+    _target_: llmbc.dataset.pointmaze_lowdim_dataset.PointmazeLowdimDataset
+    data_path: datasets/maze2d-medium-v0.pt
+    data_path2: datasets/maze2d-medium-v0.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: 0.02
+    dummy_normalizer: false
+  instructor:
+    _target_: llmbc.translator.instructor.pointmaze_instructor.maze2d_medium_v0_instructor.Maze2dMediumV0Instructor
+llm:
+  name: HuggingFaceTB/SmolLM2-135M-Instruct
+  model_name: SmolLM2-135M-Instruct
+  config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+  causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+  use_quantization: false
+  use_joint_mlp_projector: true
+  llm_mode: ete-finetuned
+  finetune_mode: orig
+  checkpoint: data/outputs/2025.11.04/14.51.54_train_llm_lowdim_maze2d-medium-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-maze2d-medium-v0/checkpoint-11850
+  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}

2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/hydra.yaml

@@ -0,0 +1,156 @@
+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:
+    - policy.loss_llm_weight=1.0e-4
+    - training.seed=45
+  job:
+    name: train
+    chdir: null
+    override_dirname: policy.loss_llm_weight=1.0e-4,training.seed=45
+    id: ???
+    num: ???
+    config_name: llmdp_maze2d-medium-v0.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/LLM-BC/data/outputs/2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0
+    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

2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/.hydra/overrides.yaml

@@ -0,0 +1,2 @@
+- policy.loss_llm_weight=1.0e-4
+- training.seed=45

2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0620-test_success_rate=0.900.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61e4af4c75873c7afd825938b6d0ced046798bb0365196d07ac9a970f93d2179
+size 1039753586

2025.12.20/17.23.30_train_llm_diffusion_unet_lowdim_maze2d-medium-v0/checkpoints/epoch=0670-test_success_rate=0.920.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:425d5acb3c09db767c0653bf26372a0505f0b9e370b26f93af32e59d1166a00e
+size 1039753586