Upload folder using huggingface_hub

.gitattributes

@@ -43,3 +43,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 2025.12.16/16.18.02_train_llmbc_lowdim_box-close-v2/wandb/run-20251216_161817-agd5zxwx/run-agd5zxwx.wandb filter=lfs diff=lfs merge=lfs -text
 2025.12.16/16.18.07_train_llmbc_lowdim_box-close-v2/logs.json.txt filter=lfs diff=lfs merge=lfs -text
 2025.12.16/16.18.07_train_llmbc_lowdim_box-close-v2/wandb/run-20251216_161825-iyphymsd/run-iyphymsd.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/run-juft93rb.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/run-z04notzj.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-v0/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-v0/wandb/run-20251217_042650-5vils759/run-5vils759.wandb filter=lfs diff=lfs merge=lfs -text
+2025.12.17/04.29.18_train_llm_diffusion_unet_lowdim_parking-v0/logs.json.txt filter=lfs diff=lfs merge=lfs -text
+2025.12.17/04.29.18_train_llm_diffusion_unet_lowdim_parking-v0/wandb/run-20251217_042930-etljn2u5/run-etljn2u5.wandb filter=lfs diff=lfs merge=lfs -text

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/.hydra/config.yaml

@@ -0,0 +1,191 @@
+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: 0.01
+  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: PegInsertionSide-v1
+  obs_dim: 43
+  action_dim: 8
+  env_runner:
+    _target_: llmbc.env_runner.maniskill_lowdim_runner.ManiskillLowdimRunner
+    env_name: llf-maniskill-PegInsertionSide-v1
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    visual: false
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 100
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    discount: 0.99
+  dataset:
+    _target_: llmbc.dataset.maniskill_lowdim_dataset.ManiskillLowdimDataset
+    data_path: datasets/PegInsertionSide-v1.pt
+    data_path2: datasets/PegInsertionSide-v1.pt
+    horizon: ${horizon}
+    pad_before: ${eval:'${n_obs_steps}-1'}
+    pad_after: ${eval:'${n_action_steps}-1'}
+    use_manual_normalizer: false
+    val_ratio: 0.02
+    dummy_normalizer: false
+  instructor:
+    _target_: llmbc.translator.instructor.maniskill_instructor.PegInsertionSide_v1_instructor.PegInsertionSideV1Instructor
+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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070
+  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}

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+hydra:
+  run:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  sweep:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - training.seed=45
+  job:
+    name: train
+    chdir: null
+    override_dirname: training.seed=45
+    id: ???
+    num: ???
+    config_name: llmdp_PegInsertionSide-v1.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /root/workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /root/workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1
+    choices:
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- training.seed=45

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/train.log

@@ -0,0 +1,9 @@
+[2025-12-17 02:11:08,360][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2025-12-17 02:11:08,360][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2025-12-17 02:11:08,360][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2025-12-17 02:11:10,305][datasets][INFO] - PyTorch version 2.2.2 available.
+[2025-12-17 02:11:10,306][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2025-12-17 02:11:10,307][datasets][INFO] - JAX version 0.4.30 available.
+[2025-12-17 02:11:11,782][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.550938e+07
+[2025-12-17 02:11:17,816][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
+[2025-12-17 02:11:17,821][absl][INFO] - MuJoCo library version is: 2.3.7

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/debug-internal.log

@@ -0,0 +1,8 @@
+{"time":"2025-12-17T02:11:18.620405938Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-17T02:11:18.62041675Z","level":"INFO","msg":"created symlink","path":"/root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug-core.log"}
+{"time":"2025-12-17T02:11:18.730784724Z","level":"INFO","msg":"created new stream","id":"juft93rb"}
+{"time":"2025-12-17T02:11:18.730817952Z","level":"INFO","msg":"stream: started","id":"juft93rb"}
+{"time":"2025-12-17T02:11:18.73084374Z","level":"INFO","msg":"sender: started","stream_id":"juft93rb"}
+{"time":"2025-12-17T02:11:18.730829993Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"juft93rb"}}
+{"time":"2025-12-17T02:11:18.730837302Z","level":"INFO","msg":"handler: started","stream_id":{"value":"juft93rb"}}
+{"time":"2025-12-17T02:11:19.406011902Z","level":"INFO","msg":"Starting system monitor"}

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/debug.log

@@ -0,0 +1,26 @@
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Configure stats pid to 23932
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Loading settings from /root/workspace/LLM-BC/wandb/settings
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/root/workspace/LLM-BC/train.py', 'program': '/root/workspace/LLM-BC/./train.py'}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:_log_setup():533] Logging user logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug.log
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:_log_setup():534] Logging internal logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug-internal.log
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:init():619] calling init triggers
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [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': 43, 'action_dim': 8, 'task_name': 'PegInsertionSide-v1', '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': 8, 'local_cond_dim': None, 'global_cond_dim': 43, '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': 43, 'action_dim': 8, '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': 'PegInsertionSide-v1', '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.11.06_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 43, 'action_dim': 8, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.01, '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': 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': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'PegInsertionSide-v1-training', 'resume': True, 'mode': 'online', 'name': '2025.12.17-02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'tags': ['train_llm_diffusion_unet_lowdim', 'PegInsertionSide-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'wandb_name_base': '2025.12.17-02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}, 'task': {'name': 'PegInsertionSide-v1', 'obs_dim': 43, 'action_dim': 8, 'env_runner': {'_target_': 'llmbc.env_runner.maniskill_lowdim_runner.ManiskillLowdimRunner', 'env_name': 'llf-maniskill-PegInsertionSide-v1', 'instruction_type': 'b', 'feedback_type': ['hp', 'hn'], 'visual': False, 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 100, 'n_obs_steps': 1, 'n_action_steps': 1, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.maniskill_lowdim_dataset.ManiskillLowdimDataset', 'data_path': 'datasets/PegInsertionSide-v1.pt', 'data_path2': 'datasets/PegInsertionSide-v1.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.maniskill_instructor.PegInsertionSide_v1_instructor.PegInsertionSideV1Instructor'}}, '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.11.06_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:init():669] starting backend
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:init():673] sending inform_init request
+2025-12-17 02:11:18,619 INFO    MainThread:23932 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-17 02:11:18,619 INFO    MainThread:23932 [wandb_init.py:init():686] backend started and connected
+2025-12-17 02:11:18,626 INFO    MainThread:23932 [wandb_init.py:init():781] updated telemetry
+2025-12-17 02:11:18,657 INFO    MainThread:23932 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-17 02:11:19,403 INFO    MainThread:23932 [wandb_init.py:init():867] starting run threads in backend
+2025-12-17 02:11:19,745 INFO    MainThread:23932 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-17 02:11:19,745 INFO    MainThread:23932 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-17 02:11:19,745 INFO    MainThread:23932 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-17 02:11:19,746 INFO    MainThread:23932 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-17 02:11:19,747 INFO    MainThread:23932 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-17 02:11:19,747 INFO    MainThread:23932 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/files/output.log

@@ -0,0 +1 @@
+                                                                                                                    

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/files/requirements.txt

@@ -0,0 +1,861 @@
+aiohappyeyeballs==2.6.1
+antlr4-python3-runtime==4.9.3
+asciitree==0.3.3
+async-timeout==5.0.1
+attrs==25.4.0
+certifi==2025.10.5
+charset-normalizer==3.4.4
+colorama==0.4.6
+dill==0.3.8
+docstring-parser==0.16
+eval_type_backport==0.2.2
+fasteners==0.20
+filelock==3.17.0
+frozenlist==1.8.0
+fsspec==2024.3.1
+gmpy2==2.2.1
+hf-xet==1.1.8
+idna==3.11
+llvmlite==0.39.1
+MarkupSafe==3.0.2
+mdurl==0.1.2
+mkl-service==2.4.0
+mpmath==1.3.0
+msgpack==1.1.1
+networkx==3.2.1
+packaging==25.0
+propcache==0.3.1
+psutil==7.0.0
+pycparser==2.23
+Pygments==2.19.1
+PyOpenGL==3.1.10
+PySocks==1.7.1
+tzdata==2025.2
+xxhash==3.5.0
+pytz==2025.2
+PyYAML==6.0.2
+regex==2025.9.1
+safetensors==0.5.3
+setuptools==80.9.0
+shtab==1.7.2
+six==1.17.0
+tqdm==4.67.1
+tensorflow==2.15.1
+gym==0.23.1
+uvloop==0.22.1
+wheel==0.45.1
+zipp==3.21.0
+requests-oauthlib==2.0.0
+Markdown==3.9
+google-auth-oauthlib==1.2.3
+tensorboard==2.15.2
+Jinja2==3.1.6
+markdown-it-py==2.2.0
+multiprocess==0.70.15
+numpy==1.23.5
+pip==23.3.2
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+brotlicffi==1.0.9.2
+hydra-core==1.2.0
+multidict==6.7.0
+pillow==11.3.0
+pycares==4.10.0
+rich==14.2.0
+sympy==1.14.0
+typeguard==4.4.4
+aiodns==3.5.0
+urllib3==2.5.0
+yarl==1.18.0
+aiohttp==3.12.15
+pyarrow==21.0.0
+astunparse==1.6.3
+pyarrow-hotfix==0.7
+tokenizers==0.21.0
+av==10.0.0
+einops==0.4.1
+Bottleneck==1.4.2
+mkl_fft==1.3.11
+mkl_random==1.2.8
+numba==0.56.4
+numcodecs==0.12.1
+numexpr==2.10.1
+pandas==2.3.3
+zarr==2.12.0
+datasets==2.19.0
+transformers==4.47.1
+accelerate==1.0.1
+peft==0.14.0
+trl==0.11.4
+torch==2.2.2
+tabulate==0.9.0
+sentencepiece==0.2.0
+pyperclip==1.11.0
+PyOpenGL-accelerate==3.1.10
+pydub==0.25.1
+py-cpuinfo==9.0.0
+pure_eval==0.2.3
+ptyprocess==0.7.0
+nvidia-ml-py==13.590.44
+hjson==3.1.0
+gym-notices==0.1.0
+glfw==2.0.0
+Farama-Notifications==0.0.4
+websockets==15.0.1
+wcwidth==0.2.14
+uc-micro-py==1.0.3
+typing-inspection==0.4.2
+triton==2.2.0
+trimesh==4.10.1
+transforms3d==0.4.2
+traitlets==5.14.3
+threadpoolctl==3.6.0
+tenacity==9.1.2
+sniffio==1.3.1
+smmap==5.0.2
+setproctitle==1.3.7
+sentry-sdk==2.47.0
+semantic-version==2.10.0
+scipy==1.13.1
+rpds-py==0.27.1
+python-multipart==0.0.20
+pyparsing==3.2.5
+pynvml==13.0.1
+pydantic_core==2.41.5
+pycryptodomex==3.23.0
+pyasn1==0.6.1
+google-pasta==0.2.0
+platformdirs==4.4.0
+pexpect==4.9.0
+parso==0.8.5
+orjson==3.11.5
+opencv-python==4.11.0.86
+stack-data==0.6.3
+omegaconf==2.2.1
+nvidia-nvtx-cu12==12.1.105
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-nccl-cu12==2.19.3
+nvidia-curand-cu12==10.3.2.106
+nvidia-cufft-cu12==11.0.2.54
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cublas-cu12==12.1.3.1
+ninja==1.13.0
+narwhals==2.13.0
+lxml==6.0.2
+kiwisolver==1.4.7
+joblib==1.5.3
+jiter==0.12.0
+imageio-ffmpeg==0.6.0
+ImageIO==2.37.2
+h5py==3.14.0
+h11==0.16.0
+fonttools==4.60.2
+ffmpy==1.0.0
+fast_kinematics==0.2.2
+executing==2.2.1
+exceptiongroup==1.3.1
+docker-pycreds==0.4.0
+distro==1.9.0
+decorator==5.2.1
+dacite==1.9.2
+cycler==0.12.1
+contourpy==1.3.0
+cloudpickle==3.1.2
+click==8.1.8
+cachetools==6.2.4
+asttokens==3.0.1
+annotated-types==0.7.0
+annotated-doc==0.0.4
+aiofiles==25.1.0
+absl-py==2.3.1
+uvicorn==0.38.0
+tiktoken==0.8.0
+scikit-learn==1.6.1
+sapien==3.0.0b1
+rsa==4.9.1
+referencing==0.36.2
+pydantic==2.12.5
+pyasn1_modules==0.4.2
+prompt_toolkit==3.0.52
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-cudnn-cu12==8.9.2.26
+nltk==3.9.2
+mujoco==2.3.7
+mdit-py-plugins==0.3.3
+matplotlib-inline==0.2.1
+matplotlib==3.7.5
+linkify-it-py==2.0.3
+jedi==0.19.2
+huggingface-hub==0.26.2
+httpcore==1.0.9
+gymnasium==0.29.1
+grpcio==1.76.0
+gitdb==4.0.12
+blobfile==3.0.0
+anyio==4.12.0
+tyro==0.9.1
+toppra==0.6.3
+starlette==0.49.3
+seaborn==0.13.2
+rouge_score==0.1.2
+pynndescent==0.5.13
+nvidia-cusolver-cu12==11.4.5.107
+jsonschema-specifications==2025.9.1
+ipython==8.18.1
+httpx==0.28.1
+google-auth==2.41.1
+GitPython==3.1.45
+diffusers==0.31.0
+wandb==0.18.6
+umap-learn==0.5.9.post2
+openai==2.8.1
+mplib==0.1.1
+jsonschema==4.25.1
+gradio_client==0.2.9
+google-genai==1.47.0
+fastapi==0.124.4
+torchvision==0.17.2
+torchaudio==2.2.2
+sentence-transformers==3.2.1
+pytorch-seed==0.2.0
+deepspeed==0.16.1
+bitsandbytes==0.45.0
+altair==6.0.0
+timm==1.0.22
+gradio==3.36.1
+evaluate==0.4.3
+arm_pytorch_utilities==0.4.3
+tensorflow-io-gcs-filesystem==0.37.1
+tensorflow-estimator==2.15.0
+tensorboard-data-server==0.7.2
+tf-agents==0.19.0
+protobuf==4.25.8
+parse==1.19.1
+dm-tree==0.1.8
+termcolor==3.1.0
+oauthlib==3.3.1
+requests==2.32.0
+opt_einsum==3.4.0
+labmaze==1.0.6
+importlib-resources==5.13.0
+wrapt==1.14.2
+importlib-metadata==5.2.0
+FLAML==2.3.6
+dm-env==1.6
+diskcache==5.6.3
+Cython==0.29.37
+mujoco-py==2.1.2.14
+jaxlib==0.4.30
+dm-control==1.0.14
+cmudict==1.0.13
+syllables==1.0.9
+metaworld==2.0.0
+jax==0.4.30
+highway-env==1.9.1
+gym_bandits==0.0.2
+d4rl==1.1
+pyautogen==0.1.0
+pybullet==3.2.6
+Werkzeug==3.1.4
+flatbuffers==25.9.23
+pytorch-kinematics==0.7.5
+pygame==2.6.1
+gin-config==0.5.0
+mani_skill==3.0.0b20
+tifffile==2024.8.30
+Shapely==1.8.4
+PyWavelets==1.6.0
+cffi==1.17.1
+scikit-image==0.19.3
+pymunk==6.2.1
+ml-dtypes==0.3.2
+tensorflow-probability==0.23.0
+stable-baselines3==2.2.1
+keras==2.15.0
+gast==0.7.0
+libclang==18.1.1
+typing_extensions==4.15.0
+llfbench==0.1.0
+llmbc==0.0.0
+aiohappyeyeballs==2.6.1
+antlr4-python3-runtime==4.9.3
+asciitree==0.3.3
+async-timeout==5.0.1
+attrs==25.4.0
+certifi==2025.10.5
+charset-normalizer==3.4.4
+colorama==0.4.6
+dill==0.3.8
+docstring-parser==0.16
+eval_type_backport==0.2.2
+fasteners==0.20
+filelock==3.17.0
+frozenlist==1.8.0
+fsspec==2024.3.1
+gmpy2==2.2.1
+hf-xet==1.1.8
+idna==3.11
+llvmlite==0.39.1
+MarkupSafe==3.0.2
+mdurl==0.1.2
+mkl-service==2.4.0
+mpmath==1.3.0
+msgpack==1.1.1
+networkx==3.2.1
+packaging==25.0
+propcache==0.3.1
+psutil==7.0.0
+pycparser==2.23
+Pygments==2.19.1
+PyOpenGL==3.1.10
+PySocks==1.7.1
+tzdata==2025.2
+xxhash==3.5.0
+pytz==2025.2
+PyYAML==6.0.2
+regex==2025.9.1
+safetensors==0.5.3
+setuptools==80.9.0
+shtab==1.7.2
+six==1.17.0
+tqdm==4.67.1
+tensorflow==2.15.1
+gym==0.23.1
+uvloop==0.22.1
+wheel==0.45.1
+zipp==3.21.0
+requests-oauthlib==2.0.0
+Markdown==3.9
+google-auth-oauthlib==1.2.3
+tensorboard==2.15.2
+Jinja2==3.1.6
+markdown-it-py==2.2.0
+multiprocess==0.70.15
+numpy==1.23.5
+pip==23.3.2
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+brotlicffi==1.0.9.2
+hydra-core==1.2.0
+multidict==6.7.0
+pillow==11.3.0
+pycares==4.10.0
+rich==14.2.0
+sympy==1.14.0
+typeguard==4.4.4
+aiodns==3.5.0
+urllib3==2.5.0
+yarl==1.18.0
+aiohttp==3.12.15
+pyarrow==21.0.0
+astunparse==1.6.3
+pyarrow-hotfix==0.7
+tokenizers==0.21.0
+av==10.0.0
+einops==0.4.1
+Bottleneck==1.4.2
+mkl_fft==1.3.11
+mkl_random==1.2.8
+numba==0.56.4
+numcodecs==0.12.1
+numexpr==2.10.1
+pandas==2.3.3
+zarr==2.12.0
+datasets==2.19.0
+transformers==4.47.1
+accelerate==1.0.1
+peft==0.14.0
+trl==0.11.4
+torch==2.2.2
+tabulate==0.9.0
+sentencepiece==0.2.0
+pyperclip==1.11.0
+PyOpenGL-accelerate==3.1.10
+pydub==0.25.1
+py-cpuinfo==9.0.0
+pure_eval==0.2.3
+ptyprocess==0.7.0
+nvidia-ml-py==13.590.44
+hjson==3.1.0
+gym-notices==0.1.0
+glfw==2.0.0
+Farama-Notifications==0.0.4
+websockets==15.0.1
+wcwidth==0.2.14
+uc-micro-py==1.0.3
+typing-inspection==0.4.2
+triton==2.2.0
+trimesh==4.10.1
+transforms3d==0.4.2
+traitlets==5.14.3
+threadpoolctl==3.6.0
+tenacity==9.1.2
+sniffio==1.3.1
+smmap==5.0.2
+setproctitle==1.3.7
+sentry-sdk==2.47.0
+semantic-version==2.10.0
+scipy==1.13.1
+rpds-py==0.27.1
+python-multipart==0.0.20
+pyparsing==3.2.5
+pynvml==13.0.1
+pydantic_core==2.41.5
+pycryptodomex==3.23.0
+pyasn1==0.6.1
+google-pasta==0.2.0
+platformdirs==4.4.0
+pexpect==4.9.0
+parso==0.8.5
+orjson==3.11.5
+opencv-python==4.11.0.86
+stack-data==0.6.3
+omegaconf==2.2.1
+nvidia-nvtx-cu12==12.1.105
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-nccl-cu12==2.19.3
+nvidia-curand-cu12==10.3.2.106
+nvidia-cufft-cu12==11.0.2.54
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cublas-cu12==12.1.3.1
+ninja==1.13.0
+narwhals==2.13.0
+lxml==6.0.2
+kiwisolver==1.4.7
+joblib==1.5.3
+jiter==0.12.0
+imageio-ffmpeg==0.6.0
+ImageIO==2.37.2
+h5py==3.14.0
+h11==0.16.0
+fonttools==4.60.2
+ffmpy==1.0.0
+fast_kinematics==0.2.2
+executing==2.2.1
+exceptiongroup==1.3.1
+docker-pycreds==0.4.0
+distro==1.9.0
+decorator==5.2.1
+dacite==1.9.2
+cycler==0.12.1
+contourpy==1.3.0
+cloudpickle==3.1.2
+click==8.1.8
+cachetools==6.2.4
+asttokens==3.0.1
+annotated-types==0.7.0
+annotated-doc==0.0.4
+aiofiles==25.1.0
+absl-py==2.3.1
+uvicorn==0.38.0
+tiktoken==0.8.0
+scikit-learn==1.6.1
+sapien==3.0.0b1
+rsa==4.9.1
+referencing==0.36.2
+pydantic==2.12.5
+pyasn1_modules==0.4.2
+prompt_toolkit==3.0.52
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-cudnn-cu12==8.9.2.26
+nltk==3.9.2
+mujoco==2.3.7
+mdit-py-plugins==0.3.3
+matplotlib-inline==0.2.1
+matplotlib==3.7.5
+linkify-it-py==2.0.3
+jedi==0.19.2
+huggingface-hub==0.26.2
+httpcore==1.0.9
+gymnasium==0.29.1
+grpcio==1.76.0
+gitdb==4.0.12
+blobfile==3.0.0
+anyio==4.12.0
+tyro==0.9.1
+toppra==0.6.3
+starlette==0.49.3
+seaborn==0.13.2
+rouge_score==0.1.2
+pynndescent==0.5.13
+nvidia-cusolver-cu12==11.4.5.107
+jsonschema-specifications==2025.9.1
+ipython==8.18.1
+httpx==0.28.1
+google-auth==2.41.1
+GitPython==3.1.45
+diffusers==0.31.0
+wandb==0.18.6
+umap-learn==0.5.9.post2
+openai==2.8.1
+mplib==0.1.1
+jsonschema==4.25.1
+gradio_client==0.2.9
+google-genai==1.47.0
+fastapi==0.124.4
+torchvision==0.17.2
+torchaudio==2.2.2
+sentence-transformers==3.2.1
+pytorch-seed==0.2.0
+deepspeed==0.16.1
+bitsandbytes==0.45.0
+altair==6.0.0
+timm==1.0.22
+gradio==3.36.1
+evaluate==0.4.3
+arm_pytorch_utilities==0.4.3
+tensorflow-io-gcs-filesystem==0.37.1
+tensorflow-estimator==2.15.0
+tensorboard-data-server==0.7.2
+tf-agents==0.19.0
+protobuf==4.25.8
+parse==1.19.1
+dm-tree==0.1.8
+termcolor==3.1.0
+oauthlib==3.3.1
+requests==2.32.0
+opt_einsum==3.4.0
+labmaze==1.0.6
+importlib-resources==5.13.0
+wrapt==1.14.2
+importlib-metadata==5.2.0
+FLAML==2.3.6
+dm-env==1.6
+diskcache==5.6.3
+Cython==0.29.37
+mujoco-py==2.1.2.14
+jaxlib==0.4.30
+dm-control==1.0.14
+cmudict==1.0.13
+syllables==1.0.9
+metaworld==2.0.0
+jax==0.4.30
+highway-env==1.9.1
+gym_bandits==0.0.2
+d4rl==1.1
+pyautogen==0.1.0
+pybullet==3.2.6
+Werkzeug==3.1.4
+flatbuffers==25.9.23
+pytorch-kinematics==0.7.5
+pygame==2.6.1
+gin-config==0.5.0
+mani_skill==3.0.0b20
+tifffile==2024.8.30
+Shapely==1.8.4
+PyWavelets==1.6.0
+cffi==1.17.1
+scikit-image==0.19.3
+pymunk==6.2.1
+ml-dtypes==0.3.2
+tensorflow-probability==0.23.0
+stable-baselines3==2.2.1
+keras==2.15.0
+gast==0.7.0
+libclang==18.1.1
+typing_extensions==4.15.0
+llfbench==0.1.0
+llmbc==0.0.0
+llmbc==0.0.0
+llmbc==0.0.0
+aiohappyeyeballs==2.6.1
+antlr4-python3-runtime==4.9.3
+asciitree==0.3.3
+async-timeout==5.0.1
+attrs==25.4.0
+certifi==2025.10.5
+charset-normalizer==3.4.4
+colorama==0.4.6
+dill==0.3.8
+docstring-parser==0.16
+eval_type_backport==0.2.2
+fasteners==0.20
+filelock==3.17.0
+frozenlist==1.8.0
+fsspec==2024.3.1
+gmpy2==2.2.1
+hf-xet==1.1.8
+idna==3.11
+llvmlite==0.39.1
+MarkupSafe==3.0.2
+mdurl==0.1.2
+mkl-service==2.4.0
+mpmath==1.3.0
+msgpack==1.1.1
+networkx==3.2.1
+packaging==25.0
+propcache==0.3.1
+psutil==7.0.0
+pycparser==2.23
+Pygments==2.19.1
+PyOpenGL==3.1.10
+PySocks==1.7.1
+tzdata==2025.2
+xxhash==3.5.0
+pytz==2025.2
+PyYAML==6.0.2
+regex==2025.9.1
+safetensors==0.5.3
+setuptools==80.9.0
+shtab==1.7.2
+six==1.17.0
+tqdm==4.67.1
+tensorflow==2.15.1
+gym==0.23.1
+uvloop==0.22.1
+wheel==0.45.1
+zipp==3.21.0
+requests-oauthlib==2.0.0
+Markdown==3.9
+google-auth-oauthlib==1.2.3
+tensorboard==2.15.2
+Jinja2==3.1.6
+markdown-it-py==2.2.0
+multiprocess==0.70.15
+numpy==1.23.5
+pip==23.3.2
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+brotlicffi==1.0.9.2
+hydra-core==1.2.0
+multidict==6.7.0
+pillow==11.3.0
+pycares==4.10.0
+rich==14.2.0
+sympy==1.14.0
+typeguard==4.4.4
+aiodns==3.5.0
+urllib3==2.5.0
+yarl==1.18.0
+aiohttp==3.12.15
+pyarrow==21.0.0
+astunparse==1.6.3
+pyarrow-hotfix==0.7
+tokenizers==0.21.0
+av==10.0.0
+einops==0.4.1
+Bottleneck==1.4.2
+mkl_fft==1.3.11
+mkl_random==1.2.8
+numba==0.56.4
+numcodecs==0.12.1
+numexpr==2.10.1
+pandas==2.3.3
+zarr==2.12.0
+datasets==2.19.0
+transformers==4.47.1
+accelerate==1.0.1
+peft==0.14.0
+trl==0.11.4
+torch==2.2.2
+tabulate==0.9.0
+sentencepiece==0.2.0
+pyperclip==1.11.0
+PyOpenGL-accelerate==3.1.10
+pydub==0.25.1
+py-cpuinfo==9.0.0
+pure_eval==0.2.3
+ptyprocess==0.7.0
+nvidia-ml-py==13.590.44
+hjson==3.1.0
+gym-notices==0.1.0
+glfw==2.0.0
+Farama-Notifications==0.0.4
+websockets==15.0.1
+wcwidth==0.2.14
+uc-micro-py==1.0.3
+typing-inspection==0.4.2
+triton==2.2.0
+trimesh==4.10.1
+transforms3d==0.4.2
+traitlets==5.14.3
+threadpoolctl==3.6.0
+tenacity==9.1.2
+sniffio==1.3.1
+smmap==5.0.2
+setproctitle==1.3.7
+sentry-sdk==2.47.0
+semantic-version==2.10.0
+scipy==1.13.1
+rpds-py==0.27.1
+python-multipart==0.0.20
+pyparsing==3.2.5
+pynvml==13.0.1
+pydantic_core==2.41.5
+pycryptodomex==3.23.0
+pyasn1==0.6.1
+google-pasta==0.2.0
+platformdirs==4.4.0
+pexpect==4.9.0
+parso==0.8.5
+orjson==3.11.5
+opencv-python==4.11.0.86
+stack-data==0.6.3
+omegaconf==2.2.1
+nvidia-nvtx-cu12==12.1.105
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-nccl-cu12==2.19.3
+nvidia-curand-cu12==10.3.2.106
+nvidia-cufft-cu12==11.0.2.54
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cublas-cu12==12.1.3.1
+ninja==1.13.0
+narwhals==2.13.0
+lxml==6.0.2
+kiwisolver==1.4.7
+joblib==1.5.3
+jiter==0.12.0
+imageio-ffmpeg==0.6.0
+ImageIO==2.37.2
+h5py==3.14.0
+h11==0.16.0
+fonttools==4.60.2
+ffmpy==1.0.0
+fast_kinematics==0.2.2
+executing==2.2.1
+exceptiongroup==1.3.1
+docker-pycreds==0.4.0
+distro==1.9.0
+decorator==5.2.1
+dacite==1.9.2
+cycler==0.12.1
+contourpy==1.3.0
+cloudpickle==3.1.2
+click==8.1.8
+cachetools==6.2.4
+asttokens==3.0.1
+annotated-types==0.7.0
+annotated-doc==0.0.4
+aiofiles==25.1.0
+absl-py==2.3.1
+uvicorn==0.38.0
+tiktoken==0.8.0
+scikit-learn==1.6.1
+sapien==3.0.0b1
+rsa==4.9.1
+referencing==0.36.2
+pydantic==2.12.5
+pyasn1_modules==0.4.2
+prompt_toolkit==3.0.52
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-cudnn-cu12==8.9.2.26
+nltk==3.9.2
+mujoco==2.3.7
+mdit-py-plugins==0.3.3
+matplotlib-inline==0.2.1
+matplotlib==3.7.5
+linkify-it-py==2.0.3
+jedi==0.19.2
+huggingface-hub==0.26.2
+httpcore==1.0.9
+gymnasium==0.29.1
+grpcio==1.76.0
+gitdb==4.0.12
+blobfile==3.0.0
+anyio==4.12.0
+tyro==0.9.1
+toppra==0.6.3
+starlette==0.49.3
+seaborn==0.13.2
+rouge_score==0.1.2
+pynndescent==0.5.13
+nvidia-cusolver-cu12==11.4.5.107
+jsonschema-specifications==2025.9.1
+ipython==8.18.1
+httpx==0.28.1
+google-auth==2.41.1
+GitPython==3.1.45
+diffusers==0.31.0
+wandb==0.18.6
+umap-learn==0.5.9.post2
+openai==2.8.1
+mplib==0.1.1
+jsonschema==4.25.1
+gradio_client==0.2.9
+google-genai==1.47.0
+fastapi==0.124.4
+torchvision==0.17.2
+torchaudio==2.2.2
+sentence-transformers==3.2.1
+pytorch-seed==0.2.0
+deepspeed==0.16.1
+bitsandbytes==0.45.0
+altair==6.0.0
+timm==1.0.22
+gradio==3.36.1
+evaluate==0.4.3
+arm_pytorch_utilities==0.4.3
+tensorflow-io-gcs-filesystem==0.37.1
+tensorflow-estimator==2.15.0
+tensorboard-data-server==0.7.2
+tf-agents==0.19.0
+protobuf==4.25.8
+parse==1.19.1
+dm-tree==0.1.8
+termcolor==3.1.0
+oauthlib==3.3.1
+requests==2.32.0
+opt_einsum==3.4.0
+labmaze==1.0.6
+importlib-resources==5.13.0
+wrapt==1.14.2
+importlib-metadata==5.2.0
+FLAML==2.3.6
+dm-env==1.6
+diskcache==5.6.3
+Cython==0.29.37
+mujoco-py==2.1.2.14
+jaxlib==0.4.30
+dm-control==1.0.14
+cmudict==1.0.13
+syllables==1.0.9
+metaworld==2.0.0
+jax==0.4.30
+highway-env==1.9.1
+gym_bandits==0.0.2
+d4rl==1.1
+pyautogen==0.1.0
+pybullet==3.2.6
+Werkzeug==3.1.4
+flatbuffers==25.9.23
+pytorch-kinematics==0.7.5
+pygame==2.6.1
+gin-config==0.5.0
+mani_skill==3.0.0b20
+tifffile==2024.8.30
+Shapely==1.8.4
+PyWavelets==1.6.0
+cffi==1.17.1
+scikit-image==0.19.3
+pymunk==6.2.1
+ml-dtypes==0.3.2
+tensorflow-probability==0.23.0
+stable-baselines3==2.2.1
+keras==2.15.0
+gast==0.7.0
+libclang==18.1.1
+typing_extensions==4.15.0
+llfbench==0.1.0
+llmbc==0.0.0
+autocommand==2.2.2
+backports.tarfile==1.2.0
+importlib_metadata==8.0.0
+inflect==7.3.1
+jaraco.collections==5.1.0
+jaraco.context==5.3.0
+jaraco.functools==4.0.1
+jaraco.text==3.12.1
+more-itertools==10.3.0
+packaging==24.2
+platformdirs==4.2.2
+tomli==2.0.1
+typeguard==4.3.0
+typing_extensions==4.12.2
+wheel==0.45.1
+zipp==3.19.2

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/files/wandb-metadata.json

@@ -0,0 +1,55 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.31",
+  "python": "3.9.25",
+  "startedAt": "2025-12-17T02:11:18.619741Z",
+  "args": [
+    "--config-path=./config/main_table",
+    "--config-name",
+    "llmdp_PegInsertionSide-v1.yaml",
+    "training.seed=45"
+  ],
+  "program": "/root/workspace/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "6f691a82b1eb85d65cc243f066e7db18b995dd32"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1",
+  "host": "llmbc-cuda-545d8cd8bc-9ltjf",
+  "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": "23041585184768",
+      "used": "532379566080"
+    }
+  },
+  "memory": {
+    "total": "2163619737600"
+  },
+  "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"
+}

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug-core.log

@@ -0,0 +1,8 @@
+{"time":"2025-12-17T02:11:18.027928857Z","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpi9n7re48/port-23932.txt","pid":23932,"debug":false,"disable-analytics":false}
+{"time":"2025-12-17T02:11:18.02795505Z","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2025-12-17T02:11:18.029207958Z","level":"INFO","msg":"Will exit if parent process dies.","ppid":23932}
+{"time":"2025-12-17T02:11:18.029208796Z","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":42875,"Zone":""}}
+{"time":"2025-12-17T02:11:18.222456783Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:47056"}
+{"time":"2025-12-17T02:11:18.62025719Z","level":"INFO","msg":"handleInformInit: received","streamId":"juft93rb","id":"127.0.0.1:47056"}
+{"time":"2025-12-17T02:11:18.730821425Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"juft93rb","id":"127.0.0.1:47056"}
+{"time":"2025-12-17T02:19:22.861451722Z","level":"INFO","msg":"Parent process exited, terminating service process."}

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug-internal.log

@@ -0,0 +1,8 @@
+{"time":"2025-12-17T02:11:18.620405938Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-17T02:11:18.62041675Z","level":"INFO","msg":"created symlink","path":"/root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug-core.log"}
+{"time":"2025-12-17T02:11:18.730784724Z","level":"INFO","msg":"created new stream","id":"juft93rb"}
+{"time":"2025-12-17T02:11:18.730817952Z","level":"INFO","msg":"stream: started","id":"juft93rb"}
+{"time":"2025-12-17T02:11:18.73084374Z","level":"INFO","msg":"sender: started","stream_id":"juft93rb"}
+{"time":"2025-12-17T02:11:18.730829993Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"juft93rb"}}
+{"time":"2025-12-17T02:11:18.730837302Z","level":"INFO","msg":"handler: started","stream_id":{"value":"juft93rb"}}
+{"time":"2025-12-17T02:11:19.406011902Z","level":"INFO","msg":"Starting system monitor"}

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug.log

@@ -0,0 +1,26 @@
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Configure stats pid to 23932
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Loading settings from /root/workspace/LLM-BC/wandb/settings
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/root/workspace/LLM-BC/train.py', 'program': '/root/workspace/LLM-BC/./train.py'}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:_log_setup():533] Logging user logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug.log
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:_log_setup():534] Logging internal logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/logs/debug-internal.log
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:init():619] calling init triggers
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [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': 43, 'action_dim': 8, 'task_name': 'PegInsertionSide-v1', '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': 8, 'local_cond_dim': None, 'global_cond_dim': 43, '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': 43, 'action_dim': 8, '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': 'PegInsertionSide-v1', '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.11.06_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 43, 'action_dim': 8, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.01, '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': 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': None, 'max_val_steps': None, 'tqdm_interval_sec': 1.0}, 'logging': {'project': 'PegInsertionSide-v1-training', 'resume': True, 'mode': 'online', 'name': '2025.12.17-02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'tags': ['train_llm_diffusion_unet_lowdim', 'PegInsertionSide-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'wandb_name_base': '2025.12.17-02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}, 'task': {'name': 'PegInsertionSide-v1', 'obs_dim': 43, 'action_dim': 8, 'env_runner': {'_target_': 'llmbc.env_runner.maniskill_lowdim_runner.ManiskillLowdimRunner', 'env_name': 'llf-maniskill-PegInsertionSide-v1', 'instruction_type': 'b', 'feedback_type': ['hp', 'hn'], 'visual': False, 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 100, 'n_obs_steps': 1, 'n_action_steps': 1, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.maniskill_lowdim_dataset.ManiskillLowdimDataset', 'data_path': 'datasets/PegInsertionSide-v1.pt', 'data_path2': 'datasets/PegInsertionSide-v1.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.maniskill_instructor.PegInsertionSide_v1_instructor.PegInsertionSideV1Instructor'}}, '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.11.06_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:init():669] starting backend
+2025-12-17 02:11:18,618 INFO    MainThread:23932 [wandb_init.py:init():673] sending inform_init request
+2025-12-17 02:11:18,619 INFO    MainThread:23932 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-17 02:11:18,619 INFO    MainThread:23932 [wandb_init.py:init():686] backend started and connected
+2025-12-17 02:11:18,626 INFO    MainThread:23932 [wandb_init.py:init():781] updated telemetry
+2025-12-17 02:11:18,657 INFO    MainThread:23932 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-17 02:11:19,403 INFO    MainThread:23932 [wandb_init.py:init():867] starting run threads in backend
+2025-12-17 02:11:19,745 INFO    MainThread:23932 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-17 02:11:19,745 INFO    MainThread:23932 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-17 02:11:19,745 INFO    MainThread:23932 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-17 02:11:19,746 INFO    MainThread:23932 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-17 02:11:19,747 INFO    MainThread:23932 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-17 02:11:19,747 INFO    MainThread:23932 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/root/workspace/LLM-BC/data/outputs/2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021118-juft93rb/run-juft93rb.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b288f040e608f0488669ea542db0bd4de3f3bca267a0eb487787dd2c18cac162
+size 1048576

2025.12.17/02.11.06_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/wandb-resume.json

@@ -0,0 +1 @@
+{"run_id": "juft93rb"}

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/.hydra/config.yaml

@@ -0,0 +1,191 @@
+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: 0.01
+  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: 46
+  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: PegInsertionSide-v1
+  obs_dim: 43
+  action_dim: 8
+  env_runner:
+    _target_: llmbc.env_runner.maniskill_lowdim_runner.ManiskillLowdimRunner
+    env_name: llf-maniskill-PegInsertionSide-v1
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    visual: false
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 100
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    discount: 0.99
+  dataset:
+    _target_: llmbc.dataset.maniskill_lowdim_dataset.ManiskillLowdimDataset
+    data_path: datasets/PegInsertionSide-v1.pt
+    data_path2: datasets/PegInsertionSide-v1.pt
+    horizon: ${horizon}
+    pad_before: ${eval:'${n_obs_steps}-1'}
+    pad_after: ${eval:'${n_action_steps}-1'}
+    use_manual_normalizer: false
+    val_ratio: 0.02
+    dummy_normalizer: false
+  instructor:
+    _target_: llmbc.translator.instructor.maniskill_instructor.PegInsertionSide_v1_instructor.PegInsertionSideV1Instructor
+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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070
+  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}

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+hydra:
+  run:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  sweep:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - training.seed=46
+  job:
+    name: train
+    chdir: null
+    override_dirname: training.seed=46
+    id: ???
+    num: ???
+    config_name: llmdp_PegInsertionSide-v1.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /root/workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /root/workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1
+    choices:
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- training.seed=46

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/train.log

@@ -0,0 +1,9 @@
+[2025-12-17 02:14:49,557][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2025-12-17 02:14:49,557][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2025-12-17 02:14:49,557][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2025-12-17 02:14:51,453][datasets][INFO] - PyTorch version 2.2.2 available.
+[2025-12-17 02:14:51,454][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2025-12-17 02:14:51,455][datasets][INFO] - JAX version 0.4.30 available.
+[2025-12-17 02:14:52,897][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.550938e+07
+[2025-12-17 02:14:58,911][absl][INFO] - MUJOCO_GL=osmesa, attempting to import specified OpenGL backend.
+[2025-12-17 02:14:58,917][absl][INFO] - MuJoCo library version is: 2.3.7

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/debug-internal.log

@@ -0,0 +1,11 @@
+{"time":"2025-12-17T02:14:59.714285482Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-17T02:14:59.714294531Z","level":"INFO","msg":"created symlink","path":"/root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug-core.log"}
+{"time":"2025-12-17T02:14:59.823297237Z","level":"INFO","msg":"created new stream","id":"z04notzj"}
+{"time":"2025-12-17T02:14:59.823322108Z","level":"INFO","msg":"stream: started","id":"z04notzj"}
+{"time":"2025-12-17T02:14:59.823336576Z","level":"INFO","msg":"sender: started","stream_id":"z04notzj"}
+{"time":"2025-12-17T02:14:59.823331289Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"z04notzj"}}
+{"time":"2025-12-17T02:14:59.823348393Z","level":"INFO","msg":"handler: started","stream_id":{"value":"z04notzj"}}
+{"time":"2025-12-17T02:15:00.538860394Z","level":"INFO","msg":"Starting system monitor"}
+{"time":"2025-12-17T02:21:10.734663996Z","level":"INFO","msg":"stream: closing","id":"z04notzj"}
+{"time":"2025-12-17T02:21:10.734687557Z","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2025-12-17T02:21:10.73492204Z","level":"INFO","msg":"Stopped system monitor"}

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/debug.log

@@ -0,0 +1,27 @@
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Configure stats pid to 24545
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Loading settings from /root/workspace/LLM-BC/wandb/settings
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/root/workspace/LLM-BC/train.py', 'program': '/root/workspace/LLM-BC/./train.py'}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_init.py:_log_setup():533] Logging user logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug.log
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_init.py:_log_setup():534] Logging internal logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug-internal.log
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_init.py:init():619] calling init triggers
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [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': 43, 'action_dim': 8, 'task_name': 'PegInsertionSide-v1', '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': 8, 'local_cond_dim': None, 'global_cond_dim': 43, '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': 43, 'action_dim': 8, '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': 'PegInsertionSide-v1', '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.14.48_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 43, 'action_dim': 8, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.01, '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': 46, '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': 'PegInsertionSide-v1-training', 'resume': True, 'mode': 'online', 'name': '2025.12.17-02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'tags': ['train_llm_diffusion_unet_lowdim', 'PegInsertionSide-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'wandb_name_base': '2025.12.17-02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}, 'task': {'name': 'PegInsertionSide-v1', 'obs_dim': 43, 'action_dim': 8, 'env_runner': {'_target_': 'llmbc.env_runner.maniskill_lowdim_runner.ManiskillLowdimRunner', 'env_name': 'llf-maniskill-PegInsertionSide-v1', 'instruction_type': 'b', 'feedback_type': ['hp', 'hn'], 'visual': False, 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 100, 'n_obs_steps': 1, 'n_action_steps': 1, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.maniskill_lowdim_dataset.ManiskillLowdimDataset', 'data_path': 'datasets/PegInsertionSide-v1.pt', 'data_path2': 'datasets/PegInsertionSide-v1.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.maniskill_instructor.PegInsertionSide_v1_instructor.PegInsertionSideV1Instructor'}}, '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.14.48_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [wandb_init.py:init():669] starting backend
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [wandb_init.py:init():673] sending inform_init request
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-17 02:14:59,713 INFO    MainThread:24545 [wandb_init.py:init():686] backend started and connected
+2025-12-17 02:14:59,720 INFO    MainThread:24545 [wandb_init.py:init():781] updated telemetry
+2025-12-17 02:14:59,750 INFO    MainThread:24545 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-17 02:15:00,536 INFO    MainThread:24545 [wandb_init.py:init():867] starting run threads in backend
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-17 02:15:00,879 INFO    MainThread:24545 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-17 02:15:00,879 INFO    MainThread:24545 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}
+2025-12-17 02:21:10,734 WARNING MsgRouterThr:24545 [router.py:message_loop():75] message_loop has been closed

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/files/config.yaml

@@ -0,0 +1,302 @@
+_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: 8
+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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070
+        config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+        finetune_mode: orig
+        hydra:
+            job:
+                override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+            run:
+                dir: data/outputs/2025.12.17/02.14.48_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: 2025.12.17-02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1
+        project: PegInsertionSide-v1-training
+        resume: true
+        tags:
+            - train_llm_diffusion_unet_lowdim
+            - PegInsertionSide-v1
+            - default
+model_name:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+multi_run:
+    value:
+        run_dir: data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1
+        wandb_name_base: 2025.12.17-02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1
+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: 43
+optimizer:
+    value:
+        _target_: torch.optim.AdamW
+        betas:
+            - 0.95
+            - 0.999
+        eps: 1e-08
+        lr: 0.0001
+        weight_decay: 1e-06
+output_dir:
+    value: /root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1
+past_action_visible:
+    value: false
+policy:
+    value:
+        _target_: llmbc.policy.llm_diffusion_unet_lowdim_policy.LLMDiffusionUnetLowdimPolicy
+        action_dim: 8
+        horizon: 1
+        llm_discriminator:
+            _target_: llmbc.discriminator.llm_ce_discriminator.LLMCEDiscriminator
+            llm_translator:
+                _target_: llmbc.translator.llm_translator.LLMTranslator
+                action_dim: 8
+                cfg:
+                    causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+                    checkpoint: data/outputs/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070
+                    config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+                    finetune_mode: orig
+                    hydra:
+                        job:
+                            override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+                        run:
+                            dir: data/outputs/2025.12.17/02.14.48_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: 43
+            task_id: PegInsertionSide-v1
+        loss_dp_weight: 1
+        loss_llm_weight: 0.01
+        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: 43
+            input_dim: 8
+            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: 43
+        pred_action_steps_only: false
+        reweight_llm_loss: true
+pred_action_steps_only:
+    value: false
+task:
+    value:
+        action_dim: 8
+        dataset:
+            _target_: llmbc.dataset.maniskill_lowdim_dataset.ManiskillLowdimDataset
+            data_path: datasets/PegInsertionSide-v1.pt
+            data_path2: datasets/PegInsertionSide-v1.pt
+            dummy_normalizer: false
+            horizon: 1
+            pad_after: 0
+            pad_before: 0
+            use_manual_normalizer: false
+            val_ratio: 0.02
+        env_runner:
+            _target_: llmbc.env_runner.maniskill_lowdim_runner.ManiskillLowdimRunner
+            discount: 0.99
+            env_name: llf-maniskill-PegInsertionSide-v1
+            feedback_type:
+                - hp
+                - hn
+            instruction_type: b
+            max_steps: 100
+            n_action_steps: 1
+            n_envs: 10
+            n_obs_steps: 1
+            n_test: 50
+            n_train: 10
+            visual: false
+        instructor:
+            _target_: llmbc.translator.instructor.maniskill_instructor.PegInsertionSide_v1_instructor.PegInsertionSideV1Instructor
+        name: PegInsertionSide-v1
+        obs_dim: 43
+task_name:
+    value: PegInsertionSide-v1
+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: 46
+        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

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/files/output.log

@@ -0,0 +1,52 @@
+Traceback (most recent call last):                                                                                  
+  File "/root/workspace/LLM-BC/./train.py", line 35, in <module>
+    main()
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/main.py", line 90, in decorated_main
+    _run_hydra(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 389, in _run_hydra
+    _run_app(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 452, in _run_app
+    run_and_report(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 213, in run_and_report
+    return func()
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 453, in <lambda>
+    lambda: hydra.run(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/hydra.py", line 119, in run
+    ret = run_job(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/core/utils.py", line 186, in run_job
+    ret.return_value = task_function(task_cfg)
+  File "/root/workspace/LLM-BC/./train.py", line 32, in main
+    workspace.run()
+  File "/root/workspace/LLM-BC/llmbc/workspace/train_llm_diffusion_unet_lowdim_workspace.py", line 201, in run
+    loss.backward()
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/torch/_tensor.py", line 522, in backward
+    torch.autograd.backward(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/torch/autograd/__init__.py", line 266, in backward
+    Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
+KeyboardInterrupt
+Traceback (most recent call last):
+  File "/root/workspace/LLM-BC/./train.py", line 35, in <module>
+    main()
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/main.py", line 90, in decorated_main
+    _run_hydra(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 389, in _run_hydra
+    _run_app(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 452, in _run_app
+    run_and_report(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 213, in run_and_report
+    return func()
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/utils.py", line 453, in <lambda>
+    lambda: hydra.run(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/_internal/hydra.py", line 119, in run
+    ret = run_job(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/hydra/core/utils.py", line 186, in run_job
+    ret.return_value = task_function(task_cfg)
+  File "/root/workspace/LLM-BC/./train.py", line 32, in main
+    workspace.run()
+  File "/root/workspace/LLM-BC/llmbc/workspace/train_llm_diffusion_unet_lowdim_workspace.py", line 201, in run
+    loss.backward()
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/torch/_tensor.py", line 522, in backward
+    torch.autograd.backward(
+  File "/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/torch/autograd/__init__.py", line 266, in backward
+    Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
+KeyboardInterrupt

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/files/requirements.txt

@@ -0,0 +1,861 @@
+aiohappyeyeballs==2.6.1
+antlr4-python3-runtime==4.9.3
+asciitree==0.3.3
+async-timeout==5.0.1
+attrs==25.4.0
+certifi==2025.10.5
+charset-normalizer==3.4.4
+colorama==0.4.6
+dill==0.3.8
+docstring-parser==0.16
+eval_type_backport==0.2.2
+fasteners==0.20
+filelock==3.17.0
+frozenlist==1.8.0
+fsspec==2024.3.1
+gmpy2==2.2.1
+hf-xet==1.1.8
+idna==3.11
+llvmlite==0.39.1
+MarkupSafe==3.0.2
+mdurl==0.1.2
+mkl-service==2.4.0
+mpmath==1.3.0
+msgpack==1.1.1
+networkx==3.2.1
+packaging==25.0
+propcache==0.3.1
+psutil==7.0.0
+pycparser==2.23
+Pygments==2.19.1
+PyOpenGL==3.1.10
+PySocks==1.7.1
+tzdata==2025.2
+xxhash==3.5.0
+pytz==2025.2
+PyYAML==6.0.2
+regex==2025.9.1
+safetensors==0.5.3
+setuptools==80.9.0
+shtab==1.7.2
+six==1.17.0
+tqdm==4.67.1
+tensorflow==2.15.1
+gym==0.23.1
+uvloop==0.22.1
+wheel==0.45.1
+zipp==3.21.0
+requests-oauthlib==2.0.0
+Markdown==3.9
+google-auth-oauthlib==1.2.3
+tensorboard==2.15.2
+Jinja2==3.1.6
+markdown-it-py==2.2.0
+multiprocess==0.70.15
+numpy==1.23.5
+pip==23.3.2
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+brotlicffi==1.0.9.2
+hydra-core==1.2.0
+multidict==6.7.0
+pillow==11.3.0
+pycares==4.10.0
+rich==14.2.0
+sympy==1.14.0
+typeguard==4.4.4
+aiodns==3.5.0
+urllib3==2.5.0
+yarl==1.18.0
+aiohttp==3.12.15
+pyarrow==21.0.0
+astunparse==1.6.3
+pyarrow-hotfix==0.7
+tokenizers==0.21.0
+av==10.0.0
+einops==0.4.1
+Bottleneck==1.4.2
+mkl_fft==1.3.11
+mkl_random==1.2.8
+numba==0.56.4
+numcodecs==0.12.1
+numexpr==2.10.1
+pandas==2.3.3
+zarr==2.12.0
+datasets==2.19.0
+transformers==4.47.1
+accelerate==1.0.1
+peft==0.14.0
+trl==0.11.4
+torch==2.2.2
+tabulate==0.9.0
+sentencepiece==0.2.0
+pyperclip==1.11.0
+PyOpenGL-accelerate==3.1.10
+pydub==0.25.1
+py-cpuinfo==9.0.0
+pure_eval==0.2.3
+ptyprocess==0.7.0
+nvidia-ml-py==13.590.44
+hjson==3.1.0
+gym-notices==0.1.0
+glfw==2.0.0
+Farama-Notifications==0.0.4
+websockets==15.0.1
+wcwidth==0.2.14
+uc-micro-py==1.0.3
+typing-inspection==0.4.2
+triton==2.2.0
+trimesh==4.10.1
+transforms3d==0.4.2
+traitlets==5.14.3
+threadpoolctl==3.6.0
+tenacity==9.1.2
+sniffio==1.3.1
+smmap==5.0.2
+setproctitle==1.3.7
+sentry-sdk==2.47.0
+semantic-version==2.10.0
+scipy==1.13.1
+rpds-py==0.27.1
+python-multipart==0.0.20
+pyparsing==3.2.5
+pynvml==13.0.1
+pydantic_core==2.41.5
+pycryptodomex==3.23.0
+pyasn1==0.6.1
+google-pasta==0.2.0
+platformdirs==4.4.0
+pexpect==4.9.0
+parso==0.8.5
+orjson==3.11.5
+opencv-python==4.11.0.86
+stack-data==0.6.3
+omegaconf==2.2.1
+nvidia-nvtx-cu12==12.1.105
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-nccl-cu12==2.19.3
+nvidia-curand-cu12==10.3.2.106
+nvidia-cufft-cu12==11.0.2.54
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cublas-cu12==12.1.3.1
+ninja==1.13.0
+narwhals==2.13.0
+lxml==6.0.2
+kiwisolver==1.4.7
+joblib==1.5.3
+jiter==0.12.0
+imageio-ffmpeg==0.6.0
+ImageIO==2.37.2
+h5py==3.14.0
+h11==0.16.0
+fonttools==4.60.2
+ffmpy==1.0.0
+fast_kinematics==0.2.2
+executing==2.2.1
+exceptiongroup==1.3.1
+docker-pycreds==0.4.0
+distro==1.9.0
+decorator==5.2.1
+dacite==1.9.2
+cycler==0.12.1
+contourpy==1.3.0
+cloudpickle==3.1.2
+click==8.1.8
+cachetools==6.2.4
+asttokens==3.0.1
+annotated-types==0.7.0
+annotated-doc==0.0.4
+aiofiles==25.1.0
+absl-py==2.3.1
+uvicorn==0.38.0
+tiktoken==0.8.0
+scikit-learn==1.6.1
+sapien==3.0.0b1
+rsa==4.9.1
+referencing==0.36.2
+pydantic==2.12.5
+pyasn1_modules==0.4.2
+prompt_toolkit==3.0.52
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-cudnn-cu12==8.9.2.26
+nltk==3.9.2
+mujoco==2.3.7
+mdit-py-plugins==0.3.3
+matplotlib-inline==0.2.1
+matplotlib==3.7.5
+linkify-it-py==2.0.3
+jedi==0.19.2
+huggingface-hub==0.26.2
+httpcore==1.0.9
+gymnasium==0.29.1
+grpcio==1.76.0
+gitdb==4.0.12
+blobfile==3.0.0
+anyio==4.12.0
+tyro==0.9.1
+toppra==0.6.3
+starlette==0.49.3
+seaborn==0.13.2
+rouge_score==0.1.2
+pynndescent==0.5.13
+nvidia-cusolver-cu12==11.4.5.107
+jsonschema-specifications==2025.9.1
+ipython==8.18.1
+httpx==0.28.1
+google-auth==2.41.1
+GitPython==3.1.45
+diffusers==0.31.0
+wandb==0.18.6
+umap-learn==0.5.9.post2
+openai==2.8.1
+mplib==0.1.1
+jsonschema==4.25.1
+gradio_client==0.2.9
+google-genai==1.47.0
+fastapi==0.124.4
+torchvision==0.17.2
+torchaudio==2.2.2
+sentence-transformers==3.2.1
+pytorch-seed==0.2.0
+deepspeed==0.16.1
+bitsandbytes==0.45.0
+altair==6.0.0
+timm==1.0.22
+gradio==3.36.1
+evaluate==0.4.3
+arm_pytorch_utilities==0.4.3
+tensorflow-io-gcs-filesystem==0.37.1
+tensorflow-estimator==2.15.0
+tensorboard-data-server==0.7.2
+tf-agents==0.19.0
+protobuf==4.25.8
+parse==1.19.1
+dm-tree==0.1.8
+termcolor==3.1.0
+oauthlib==3.3.1
+requests==2.32.0
+opt_einsum==3.4.0
+labmaze==1.0.6
+importlib-resources==5.13.0
+wrapt==1.14.2
+importlib-metadata==5.2.0
+FLAML==2.3.6
+dm-env==1.6
+diskcache==5.6.3
+Cython==0.29.37
+mujoco-py==2.1.2.14
+jaxlib==0.4.30
+dm-control==1.0.14
+cmudict==1.0.13
+syllables==1.0.9
+metaworld==2.0.0
+jax==0.4.30
+highway-env==1.9.1
+gym_bandits==0.0.2
+d4rl==1.1
+pyautogen==0.1.0
+pybullet==3.2.6
+Werkzeug==3.1.4
+flatbuffers==25.9.23
+pytorch-kinematics==0.7.5
+pygame==2.6.1
+gin-config==0.5.0
+mani_skill==3.0.0b20
+tifffile==2024.8.30
+Shapely==1.8.4
+PyWavelets==1.6.0
+cffi==1.17.1
+scikit-image==0.19.3
+pymunk==6.2.1
+ml-dtypes==0.3.2
+tensorflow-probability==0.23.0
+stable-baselines3==2.2.1
+keras==2.15.0
+gast==0.7.0
+libclang==18.1.1
+typing_extensions==4.15.0
+llfbench==0.1.0
+llmbc==0.0.0
+aiohappyeyeballs==2.6.1
+antlr4-python3-runtime==4.9.3
+asciitree==0.3.3
+async-timeout==5.0.1
+attrs==25.4.0
+certifi==2025.10.5
+charset-normalizer==3.4.4
+colorama==0.4.6
+dill==0.3.8
+docstring-parser==0.16
+eval_type_backport==0.2.2
+fasteners==0.20
+filelock==3.17.0
+frozenlist==1.8.0
+fsspec==2024.3.1
+gmpy2==2.2.1
+hf-xet==1.1.8
+idna==3.11
+llvmlite==0.39.1
+MarkupSafe==3.0.2
+mdurl==0.1.2
+mkl-service==2.4.0
+mpmath==1.3.0
+msgpack==1.1.1
+networkx==3.2.1
+packaging==25.0
+propcache==0.3.1
+psutil==7.0.0
+pycparser==2.23
+Pygments==2.19.1
+PyOpenGL==3.1.10
+PySocks==1.7.1
+tzdata==2025.2
+xxhash==3.5.0
+pytz==2025.2
+PyYAML==6.0.2
+regex==2025.9.1
+safetensors==0.5.3
+setuptools==80.9.0
+shtab==1.7.2
+six==1.17.0
+tqdm==4.67.1
+tensorflow==2.15.1
+gym==0.23.1
+uvloop==0.22.1
+wheel==0.45.1
+zipp==3.21.0
+requests-oauthlib==2.0.0
+Markdown==3.9
+google-auth-oauthlib==1.2.3
+tensorboard==2.15.2
+Jinja2==3.1.6
+markdown-it-py==2.2.0
+multiprocess==0.70.15
+numpy==1.23.5
+pip==23.3.2
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+brotlicffi==1.0.9.2
+hydra-core==1.2.0
+multidict==6.7.0
+pillow==11.3.0
+pycares==4.10.0
+rich==14.2.0
+sympy==1.14.0
+typeguard==4.4.4
+aiodns==3.5.0
+urllib3==2.5.0
+yarl==1.18.0
+aiohttp==3.12.15
+pyarrow==21.0.0
+astunparse==1.6.3
+pyarrow-hotfix==0.7
+tokenizers==0.21.0
+av==10.0.0
+einops==0.4.1
+Bottleneck==1.4.2
+mkl_fft==1.3.11
+mkl_random==1.2.8
+numba==0.56.4
+numcodecs==0.12.1
+numexpr==2.10.1
+pandas==2.3.3
+zarr==2.12.0
+datasets==2.19.0
+transformers==4.47.1
+accelerate==1.0.1
+peft==0.14.0
+trl==0.11.4
+torch==2.2.2
+tabulate==0.9.0
+sentencepiece==0.2.0
+pyperclip==1.11.0
+PyOpenGL-accelerate==3.1.10
+pydub==0.25.1
+py-cpuinfo==9.0.0
+pure_eval==0.2.3
+ptyprocess==0.7.0
+nvidia-ml-py==13.590.44
+hjson==3.1.0
+gym-notices==0.1.0
+glfw==2.0.0
+Farama-Notifications==0.0.4
+websockets==15.0.1
+wcwidth==0.2.14
+uc-micro-py==1.0.3
+typing-inspection==0.4.2
+triton==2.2.0
+trimesh==4.10.1
+transforms3d==0.4.2
+traitlets==5.14.3
+threadpoolctl==3.6.0
+tenacity==9.1.2
+sniffio==1.3.1
+smmap==5.0.2
+setproctitle==1.3.7
+sentry-sdk==2.47.0
+semantic-version==2.10.0
+scipy==1.13.1
+rpds-py==0.27.1
+python-multipart==0.0.20
+pyparsing==3.2.5
+pynvml==13.0.1
+pydantic_core==2.41.5
+pycryptodomex==3.23.0
+pyasn1==0.6.1
+google-pasta==0.2.0
+platformdirs==4.4.0
+pexpect==4.9.0
+parso==0.8.5
+orjson==3.11.5
+opencv-python==4.11.0.86
+stack-data==0.6.3
+omegaconf==2.2.1
+nvidia-nvtx-cu12==12.1.105
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-nccl-cu12==2.19.3
+nvidia-curand-cu12==10.3.2.106
+nvidia-cufft-cu12==11.0.2.54
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cublas-cu12==12.1.3.1
+ninja==1.13.0
+narwhals==2.13.0
+lxml==6.0.2
+kiwisolver==1.4.7
+joblib==1.5.3
+jiter==0.12.0
+imageio-ffmpeg==0.6.0
+ImageIO==2.37.2
+h5py==3.14.0
+h11==0.16.0
+fonttools==4.60.2
+ffmpy==1.0.0
+fast_kinematics==0.2.2
+executing==2.2.1
+exceptiongroup==1.3.1
+docker-pycreds==0.4.0
+distro==1.9.0
+decorator==5.2.1
+dacite==1.9.2
+cycler==0.12.1
+contourpy==1.3.0
+cloudpickle==3.1.2
+click==8.1.8
+cachetools==6.2.4
+asttokens==3.0.1
+annotated-types==0.7.0
+annotated-doc==0.0.4
+aiofiles==25.1.0
+absl-py==2.3.1
+uvicorn==0.38.0
+tiktoken==0.8.0
+scikit-learn==1.6.1
+sapien==3.0.0b1
+rsa==4.9.1
+referencing==0.36.2
+pydantic==2.12.5
+pyasn1_modules==0.4.2
+prompt_toolkit==3.0.52
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-cudnn-cu12==8.9.2.26
+nltk==3.9.2
+mujoco==2.3.7
+mdit-py-plugins==0.3.3
+matplotlib-inline==0.2.1
+matplotlib==3.7.5
+linkify-it-py==2.0.3
+jedi==0.19.2
+huggingface-hub==0.26.2
+httpcore==1.0.9
+gymnasium==0.29.1
+grpcio==1.76.0
+gitdb==4.0.12
+blobfile==3.0.0
+anyio==4.12.0
+tyro==0.9.1
+toppra==0.6.3
+starlette==0.49.3
+seaborn==0.13.2
+rouge_score==0.1.2
+pynndescent==0.5.13
+nvidia-cusolver-cu12==11.4.5.107
+jsonschema-specifications==2025.9.1
+ipython==8.18.1
+httpx==0.28.1
+google-auth==2.41.1
+GitPython==3.1.45
+diffusers==0.31.0
+wandb==0.18.6
+umap-learn==0.5.9.post2
+openai==2.8.1
+mplib==0.1.1
+jsonschema==4.25.1
+gradio_client==0.2.9
+google-genai==1.47.0
+fastapi==0.124.4
+torchvision==0.17.2
+torchaudio==2.2.2
+sentence-transformers==3.2.1
+pytorch-seed==0.2.0
+deepspeed==0.16.1
+bitsandbytes==0.45.0
+altair==6.0.0
+timm==1.0.22
+gradio==3.36.1
+evaluate==0.4.3
+arm_pytorch_utilities==0.4.3
+tensorflow-io-gcs-filesystem==0.37.1
+tensorflow-estimator==2.15.0
+tensorboard-data-server==0.7.2
+tf-agents==0.19.0
+protobuf==4.25.8
+parse==1.19.1
+dm-tree==0.1.8
+termcolor==3.1.0
+oauthlib==3.3.1
+requests==2.32.0
+opt_einsum==3.4.0
+labmaze==1.0.6
+importlib-resources==5.13.0
+wrapt==1.14.2
+importlib-metadata==5.2.0
+FLAML==2.3.6
+dm-env==1.6
+diskcache==5.6.3
+Cython==0.29.37
+mujoco-py==2.1.2.14
+jaxlib==0.4.30
+dm-control==1.0.14
+cmudict==1.0.13
+syllables==1.0.9
+metaworld==2.0.0
+jax==0.4.30
+highway-env==1.9.1
+gym_bandits==0.0.2
+d4rl==1.1
+pyautogen==0.1.0
+pybullet==3.2.6
+Werkzeug==3.1.4
+flatbuffers==25.9.23
+pytorch-kinematics==0.7.5
+pygame==2.6.1
+gin-config==0.5.0
+mani_skill==3.0.0b20
+tifffile==2024.8.30
+Shapely==1.8.4
+PyWavelets==1.6.0
+cffi==1.17.1
+scikit-image==0.19.3
+pymunk==6.2.1
+ml-dtypes==0.3.2
+tensorflow-probability==0.23.0
+stable-baselines3==2.2.1
+keras==2.15.0
+gast==0.7.0
+libclang==18.1.1
+typing_extensions==4.15.0
+llfbench==0.1.0
+llmbc==0.0.0
+llmbc==0.0.0
+llmbc==0.0.0
+aiohappyeyeballs==2.6.1
+antlr4-python3-runtime==4.9.3
+asciitree==0.3.3
+async-timeout==5.0.1
+attrs==25.4.0
+certifi==2025.10.5
+charset-normalizer==3.4.4
+colorama==0.4.6
+dill==0.3.8
+docstring-parser==0.16
+eval_type_backport==0.2.2
+fasteners==0.20
+filelock==3.17.0
+frozenlist==1.8.0
+fsspec==2024.3.1
+gmpy2==2.2.1
+hf-xet==1.1.8
+idna==3.11
+llvmlite==0.39.1
+MarkupSafe==3.0.2
+mdurl==0.1.2
+mkl-service==2.4.0
+mpmath==1.3.0
+msgpack==1.1.1
+networkx==3.2.1
+packaging==25.0
+propcache==0.3.1
+psutil==7.0.0
+pycparser==2.23
+Pygments==2.19.1
+PyOpenGL==3.1.10
+PySocks==1.7.1
+tzdata==2025.2
+xxhash==3.5.0
+pytz==2025.2
+PyYAML==6.0.2
+regex==2025.9.1
+safetensors==0.5.3
+setuptools==80.9.0
+shtab==1.7.2
+six==1.17.0
+tqdm==4.67.1
+tensorflow==2.15.1
+gym==0.23.1
+uvloop==0.22.1
+wheel==0.45.1
+zipp==3.21.0
+requests-oauthlib==2.0.0
+Markdown==3.9
+google-auth-oauthlib==1.2.3
+tensorboard==2.15.2
+Jinja2==3.1.6
+markdown-it-py==2.2.0
+multiprocess==0.70.15
+numpy==1.23.5
+pip==23.3.2
+python-dateutil==2.9.0.post0
+aiosignal==1.4.0
+brotlicffi==1.0.9.2
+hydra-core==1.2.0
+multidict==6.7.0
+pillow==11.3.0
+pycares==4.10.0
+rich==14.2.0
+sympy==1.14.0
+typeguard==4.4.4
+aiodns==3.5.0
+urllib3==2.5.0
+yarl==1.18.0
+aiohttp==3.12.15
+pyarrow==21.0.0
+astunparse==1.6.3
+pyarrow-hotfix==0.7
+tokenizers==0.21.0
+av==10.0.0
+einops==0.4.1
+Bottleneck==1.4.2
+mkl_fft==1.3.11
+mkl_random==1.2.8
+numba==0.56.4
+numcodecs==0.12.1
+numexpr==2.10.1
+pandas==2.3.3
+zarr==2.12.0
+datasets==2.19.0
+transformers==4.47.1
+accelerate==1.0.1
+peft==0.14.0
+trl==0.11.4
+torch==2.2.2
+tabulate==0.9.0
+sentencepiece==0.2.0
+pyperclip==1.11.0
+PyOpenGL-accelerate==3.1.10
+pydub==0.25.1
+py-cpuinfo==9.0.0
+pure_eval==0.2.3
+ptyprocess==0.7.0
+nvidia-ml-py==13.590.44
+hjson==3.1.0
+gym-notices==0.1.0
+glfw==2.0.0
+Farama-Notifications==0.0.4
+websockets==15.0.1
+wcwidth==0.2.14
+uc-micro-py==1.0.3
+typing-inspection==0.4.2
+triton==2.2.0
+trimesh==4.10.1
+transforms3d==0.4.2
+traitlets==5.14.3
+threadpoolctl==3.6.0
+tenacity==9.1.2
+sniffio==1.3.1
+smmap==5.0.2
+setproctitle==1.3.7
+sentry-sdk==2.47.0
+semantic-version==2.10.0
+scipy==1.13.1
+rpds-py==0.27.1
+python-multipart==0.0.20
+pyparsing==3.2.5
+pynvml==13.0.1
+pydantic_core==2.41.5
+pycryptodomex==3.23.0
+pyasn1==0.6.1
+google-pasta==0.2.0
+platformdirs==4.4.0
+pexpect==4.9.0
+parso==0.8.5
+orjson==3.11.5
+opencv-python==4.11.0.86
+stack-data==0.6.3
+omegaconf==2.2.1
+nvidia-nvtx-cu12==12.1.105
+nvidia-nvjitlink-cu12==12.9.86
+nvidia-nccl-cu12==2.19.3
+nvidia-curand-cu12==10.3.2.106
+nvidia-cufft-cu12==11.0.2.54
+nvidia-cuda-runtime-cu12==12.1.105
+nvidia-cuda-nvrtc-cu12==12.1.105
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-cublas-cu12==12.1.3.1
+ninja==1.13.0
+narwhals==2.13.0
+lxml==6.0.2
+kiwisolver==1.4.7
+joblib==1.5.3
+jiter==0.12.0
+imageio-ffmpeg==0.6.0
+ImageIO==2.37.2
+h5py==3.14.0
+h11==0.16.0
+fonttools==4.60.2
+ffmpy==1.0.0
+fast_kinematics==0.2.2
+executing==2.2.1
+exceptiongroup==1.3.1
+docker-pycreds==0.4.0
+distro==1.9.0
+decorator==5.2.1
+dacite==1.9.2
+cycler==0.12.1
+contourpy==1.3.0
+cloudpickle==3.1.2
+click==8.1.8
+cachetools==6.2.4
+asttokens==3.0.1
+annotated-types==0.7.0
+annotated-doc==0.0.4
+aiofiles==25.1.0
+absl-py==2.3.1
+uvicorn==0.38.0
+tiktoken==0.8.0
+scikit-learn==1.6.1
+sapien==3.0.0b1
+rsa==4.9.1
+referencing==0.36.2
+pydantic==2.12.5
+pyasn1_modules==0.4.2
+prompt_toolkit==3.0.52
+nvidia-cusparse-cu12==12.1.0.106
+nvidia-cudnn-cu12==8.9.2.26
+nltk==3.9.2
+mujoco==2.3.7
+mdit-py-plugins==0.3.3
+matplotlib-inline==0.2.1
+matplotlib==3.7.5
+linkify-it-py==2.0.3
+jedi==0.19.2
+huggingface-hub==0.26.2
+httpcore==1.0.9
+gymnasium==0.29.1
+grpcio==1.76.0
+gitdb==4.0.12
+blobfile==3.0.0
+anyio==4.12.0
+tyro==0.9.1
+toppra==0.6.3
+starlette==0.49.3
+seaborn==0.13.2
+rouge_score==0.1.2
+pynndescent==0.5.13
+nvidia-cusolver-cu12==11.4.5.107
+jsonschema-specifications==2025.9.1
+ipython==8.18.1
+httpx==0.28.1
+google-auth==2.41.1
+GitPython==3.1.45
+diffusers==0.31.0
+wandb==0.18.6
+umap-learn==0.5.9.post2
+openai==2.8.1
+mplib==0.1.1
+jsonschema==4.25.1
+gradio_client==0.2.9
+google-genai==1.47.0
+fastapi==0.124.4
+torchvision==0.17.2
+torchaudio==2.2.2
+sentence-transformers==3.2.1
+pytorch-seed==0.2.0
+deepspeed==0.16.1
+bitsandbytes==0.45.0
+altair==6.0.0
+timm==1.0.22
+gradio==3.36.1
+evaluate==0.4.3
+arm_pytorch_utilities==0.4.3
+tensorflow-io-gcs-filesystem==0.37.1
+tensorflow-estimator==2.15.0
+tensorboard-data-server==0.7.2
+tf-agents==0.19.0
+protobuf==4.25.8
+parse==1.19.1
+dm-tree==0.1.8
+termcolor==3.1.0
+oauthlib==3.3.1
+requests==2.32.0
+opt_einsum==3.4.0
+labmaze==1.0.6
+importlib-resources==5.13.0
+wrapt==1.14.2
+importlib-metadata==5.2.0
+FLAML==2.3.6
+dm-env==1.6
+diskcache==5.6.3
+Cython==0.29.37
+mujoco-py==2.1.2.14
+jaxlib==0.4.30
+dm-control==1.0.14
+cmudict==1.0.13
+syllables==1.0.9
+metaworld==2.0.0
+jax==0.4.30
+highway-env==1.9.1
+gym_bandits==0.0.2
+d4rl==1.1
+pyautogen==0.1.0
+pybullet==3.2.6
+Werkzeug==3.1.4
+flatbuffers==25.9.23
+pytorch-kinematics==0.7.5
+pygame==2.6.1
+gin-config==0.5.0
+mani_skill==3.0.0b20
+tifffile==2024.8.30
+Shapely==1.8.4
+PyWavelets==1.6.0
+cffi==1.17.1
+scikit-image==0.19.3
+pymunk==6.2.1
+ml-dtypes==0.3.2
+tensorflow-probability==0.23.0
+stable-baselines3==2.2.1
+keras==2.15.0
+gast==0.7.0
+libclang==18.1.1
+typing_extensions==4.15.0
+llfbench==0.1.0
+llmbc==0.0.0
+autocommand==2.2.2
+backports.tarfile==1.2.0
+importlib_metadata==8.0.0
+inflect==7.3.1
+jaraco.collections==5.1.0
+jaraco.context==5.3.0
+jaraco.functools==4.0.1
+jaraco.text==3.12.1
+more-itertools==10.3.0
+packaging==24.2
+platformdirs==4.2.2
+tomli==2.0.1
+typeguard==4.3.0
+typing_extensions==4.12.2
+wheel==0.45.1
+zipp==3.19.2

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/files/wandb-metadata.json

@@ -0,0 +1,55 @@
+{
+  "os": "Linux-4.18.0-513.24.1.el8_9.x86_64-x86_64-with-glibc2.31",
+  "python": "3.9.25",
+  "startedAt": "2025-12-17T02:14:59.713173Z",
+  "args": [
+    "--config-path=./config/main_table",
+    "--config-name",
+    "llmdp_PegInsertionSide-v1.yaml",
+    "training.seed=46"
+  ],
+  "program": "/root/workspace/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "6f691a82b1eb85d65cc243f066e7db18b995dd32"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1",
+  "host": "llmbc-cuda-545d8cd8bc-9ltjf",
+  "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": "23041585184768",
+      "used": "532379197440"
+    }
+  },
+  "memory": {
+    "total": "2163619737600"
+  },
+  "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"
+}

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/files/wandb-summary.json

@@ -0,0 +1 @@
+{"_wandb":{"runtime":371},"train_loss":0.847781777381897,"train_loss_llm":0,"_step":885,"_runtime":371.021507607,"lr":1.1200000000000001e-05,"_timestamp":1.7659380703376908e+09,"global_step":885,"epoch":0,"train_loss_dp":0.847781777381897}

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug-core.log

@@ -0,0 +1,12 @@
+{"time":"2025-12-17T02:14:59.122908307Z","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpi48_0b_0/port-24545.txt","pid":24545,"debug":false,"disable-analytics":false}
+{"time":"2025-12-17T02:14:59.122930546Z","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2025-12-17T02:14:59.123531361Z","level":"INFO","msg":"Will exit if parent process dies.","ppid":24545}
+{"time":"2025-12-17T02:14:59.123533187Z","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":35971,"Zone":""}}
+{"time":"2025-12-17T02:14:59.317792455Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:46108"}
+{"time":"2025-12-17T02:14:59.714173498Z","level":"INFO","msg":"handleInformInit: received","streamId":"z04notzj","id":"127.0.0.1:46108"}
+{"time":"2025-12-17T02:14:59.823325592Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"z04notzj","id":"127.0.0.1:46108"}
+{"time":"2025-12-17T02:21:10.734600269Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:46108"}
+{"time":"2025-12-17T02:21:10.734657102Z","level":"INFO","msg":"server is shutting down"}
+{"time":"2025-12-17T02:21:10.734655395Z","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:46108"}
+{"time":"2025-12-17T02:21:10.734703363Z","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:46108"}
+{"time":"2025-12-17T02:21:13.045892321Z","level":"INFO","msg":"Parent process exited, terminating service process."}

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug-internal.log

@@ -0,0 +1,11 @@
+{"time":"2025-12-17T02:14:59.714285482Z","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2025-12-17T02:14:59.714294531Z","level":"INFO","msg":"created symlink","path":"/root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug-core.log"}
+{"time":"2025-12-17T02:14:59.823297237Z","level":"INFO","msg":"created new stream","id":"z04notzj"}
+{"time":"2025-12-17T02:14:59.823322108Z","level":"INFO","msg":"stream: started","id":"z04notzj"}
+{"time":"2025-12-17T02:14:59.823336576Z","level":"INFO","msg":"sender: started","stream_id":"z04notzj"}
+{"time":"2025-12-17T02:14:59.823331289Z","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"z04notzj"}}
+{"time":"2025-12-17T02:14:59.823348393Z","level":"INFO","msg":"handler: started","stream_id":{"value":"z04notzj"}}
+{"time":"2025-12-17T02:15:00.538860394Z","level":"INFO","msg":"Starting system monitor"}
+{"time":"2025-12-17T02:21:10.734663996Z","level":"INFO","msg":"stream: closing","id":"z04notzj"}
+{"time":"2025-12-17T02:21:10.734687557Z","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2025-12-17T02:21:10.73492204Z","level":"INFO","msg":"Stopped system monitor"}

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug.log

@@ -0,0 +1,27 @@
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Configure stats pid to 24545
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Loading settings from /root/.config/wandb/settings
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Loading settings from /root/workspace/LLM-BC/wandb/settings
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/root/workspace/LLM-BC/train.py', 'program': '/root/workspace/LLM-BC/./train.py'}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_setup.py:_flush():79] Applying login settings: {}
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_init.py:_log_setup():533] Logging user logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug.log
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_init.py:_log_setup():534] Logging internal logs to /root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/logs/debug-internal.log
+2025-12-17 02:14:59,711 INFO    MainThread:24545 [wandb_init.py:init():619] calling init triggers
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [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': 43, 'action_dim': 8, 'task_name': 'PegInsertionSide-v1', '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': 8, 'local_cond_dim': None, 'global_cond_dim': 43, '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': 43, 'action_dim': 8, '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': 'PegInsertionSide-v1', '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.14.48_HuggingFaceTB/SmolLM2-135M-Instruct'}}}, 'obs_dim': 43, 'action_dim': 8, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1}}, 'loss_dp_weight': 1.0, 'loss_llm_weight': 0.01, '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': 46, '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': 'PegInsertionSide-v1-training', 'resume': True, 'mode': 'online', 'name': '2025.12.17-02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'tags': ['train_llm_diffusion_unet_lowdim', 'PegInsertionSide-v1', 'default'], 'id': None, 'group': None}, 'checkpoint': {'topk': {'monitor_key': 'test_success_rate', 'mode': 'max', 'k': 5, 'format_str': 'epoch={epoch:04d}-test_success_rate={test_success_rate:.3f}.ckpt'}, 'save_last_ckpt': True, 'save_last_snapshot': False}, 'multi_run': {'run_dir': 'data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1', 'wandb_name_base': '2025.12.17-02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}, 'task': {'name': 'PegInsertionSide-v1', 'obs_dim': 43, 'action_dim': 8, 'env_runner': {'_target_': 'llmbc.env_runner.maniskill_lowdim_runner.ManiskillLowdimRunner', 'env_name': 'llf-maniskill-PegInsertionSide-v1', 'instruction_type': 'b', 'feedback_type': ['hp', 'hn'], 'visual': False, 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 100, 'n_obs_steps': 1, 'n_action_steps': 1, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.maniskill_lowdim_dataset.ManiskillLowdimDataset', 'data_path': 'datasets/PegInsertionSide-v1.pt', 'data_path2': 'datasets/PegInsertionSide-v1.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'use_manual_normalizer': False, 'val_ratio': 0.02, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.maniskill_instructor.PegInsertionSide_v1_instructor.PegInsertionSideV1Instructor'}}, '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/2025.10.20/18.44.04_train_llm_lowdim_PegInsertionSide-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-PegInsertionSide-v1/checkpoint-19070', '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/2025.12.17/02.14.48_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [wandb_init.py:init():669] starting backend
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [wandb_init.py:init():673] sending inform_init request
+2025-12-17 02:14:59,712 INFO    MainThread:24545 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2025-12-17 02:14:59,713 INFO    MainThread:24545 [wandb_init.py:init():686] backend started and connected
+2025-12-17 02:14:59,720 INFO    MainThread:24545 [wandb_init.py:init():781] updated telemetry
+2025-12-17 02:14:59,750 INFO    MainThread:24545 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2025-12-17 02:15:00,536 INFO    MainThread:24545 [wandb_init.py:init():867] starting run threads in backend
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_console_start():2451] atexit reg
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2025-12-17 02:15:00,877 INFO    MainThread:24545 [wandb_run.py:_redirect():2389] Redirects installed.
+2025-12-17 02:15:00,879 INFO    MainThread:24545 [wandb_init.py:init():911] run started, returning control to user process
+2025-12-17 02:15:00,879 INFO    MainThread:24545 [wandb_run.py:_config_callback():1389] config_cb None None {'output_dir': '/root/workspace/LLM-BC/data/outputs/2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1'}
+2025-12-17 02:21:10,734 WARNING MsgRouterThr:24545 [router.py:message_loop():75] message_loop has been closed

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/run-20251217_021459-z04notzj/run-z04notzj.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d50ae7d1fbdd8e328ccc4e3d8e03036d5f541eb6ea6cb85ec0d706e016464d8a
+size 819200

2025.12.17/02.14.48_train_llm_diffusion_unet_lowdim_PegInsertionSide-v1/wandb/wandb-resume.json

@@ -0,0 +1 @@
+{"run_id": "z04notzj"}

2025.12.17/04.01.21_train_llm_diffusion_unet_lowdim_parking-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: 0.001
+  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: 42
+  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: parking-v0
+  obs_dim: 18
+  action_dim: 2
+  env_runner:
+    _target_: llmbc.env_runner.highway_lowdim_runner.HighwayLowdimRunner
+    env_name: llf-highway-parking-v0
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 80
+    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.highway_lowdim_dataset.HighwayLowdimDataset
+    data_path: datasets/parking-v0.pt
+    data_path2: datasets/parking-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.highway_instructor.parking_v0_instructor.ParkingV0Instructor
+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/2025.11.03/22.51.21_train_llm_lowdim_parking-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-parking-v0/checkpoint-5700
+  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}

2025.12.17/04.01.21_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+hydra:
+  run:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  sweep:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - policy.loss_llm_weight=1.0e-3
+  job:
+    name: train
+    chdir: null
+    override_dirname: policy.loss_llm_weight=1.0e-3
+    id: ???
+    num: ???
+    config_name: llmdp_parking-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: /root/workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /root/workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /root/workspace/LLM-BC/data/outputs/2025.12.17/04.01.21_train_llm_diffusion_unet_lowdim_parking-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

2025.12.17/04.01.21_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- policy.loss_llm_weight=1.0e-3

2025.12.17/04.01.21_train_llm_diffusion_unet_lowdim_parking-v0/train.log

@@ -0,0 +1,8 @@
+[2025-12-17 04:01:22,830][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2025-12-17 04:01:22,830][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2025-12-17 04:01:22,830][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2025-12-17 04:01:24,739][datasets][INFO] - PyTorch version 2.2.2 available.
+[2025-12-17 04:01:24,740][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2025-12-17 04:01:24,740][datasets][INFO] - JAX version 0.4.30 available.
+[2025-12-17 04:01:24,867][hydra.utils][ERROR] - Error initializing class at llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace: Error loading 'llmbc.workspace.train_llm_diffusion_unet_lowdim_workspace.TrainLLMDiffusionUnetLowdimWorkspace':
+ImportError("cannot import name 'Sentinel' from 'typing_extensions' (/root/miniconda3/envs/llm-bc/lib/python3.9/site-packages/typing_extensions.py)")

2025.12.17/04.03.03_train_llm_diffusion_unet_lowdim_parking-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: 0.001
+  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: 42
+  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: parking-v0
+  obs_dim: 18
+  action_dim: 2
+  env_runner:
+    _target_: llmbc.env_runner.highway_lowdim_runner.HighwayLowdimRunner
+    env_name: llf-highway-parking-v0
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 80
+    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.highway_lowdim_dataset.HighwayLowdimDataset
+    data_path: datasets/parking-v0.pt
+    data_path2: datasets/parking-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.highway_instructor.parking_v0_instructor.ParkingV0Instructor
+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/2025.11.03/22.51.21_train_llm_lowdim_parking-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-parking-v0/checkpoint-5700
+  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}

2025.12.17/04.03.03_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+hydra:
+  run:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  sweep:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - policy.loss_llm_weight=1.0e-3
+  job:
+    name: train
+    chdir: null
+    override_dirname: policy.loss_llm_weight=1.0e-3
+    id: ???
+    num: ???
+    config_name: llmdp_parking-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: /root/workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /root/workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /root/workspace/LLM-BC/data/outputs/2025.12.17/04.03.03_train_llm_diffusion_unet_lowdim_parking-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

2025.12.17/04.03.03_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- policy.loss_llm_weight=1.0e-3

2025.12.17/04.03.03_train_llm_diffusion_unet_lowdim_parking-v0/train.log

@@ -0,0 +1,7 @@
+[2025-12-17 04:03:05,481][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2025-12-17 04:03:05,481][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2025-12-17 04:03:05,481][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2025-12-17 04:03:07,396][datasets][INFO] - PyTorch version 2.2.2 available.
+[2025-12-17 04:03:07,397][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2025-12-17 04:03:07,398][datasets][INFO] - JAX version 0.4.30 available.
+[2025-12-17 04:03:08,817][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.514023e+07

2025.12.17/04.16.11_train_llm_diffusion_unet_lowdim_parking-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: 0.001
+  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: 42
+  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: parking-v0
+  obs_dim: 18
+  action_dim: 2
+  env_runner:
+    _target_: llmbc.env_runner.highway_lowdim_runner.HighwayLowdimRunner
+    env_name: llf-highway-parking-v0
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 80
+    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.highway_lowdim_dataset.HighwayLowdimDataset
+    data_path: datasets/parking-v0.pt
+    data_path2: datasets/parking-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.highway_instructor.parking_v0_instructor.ParkingV0Instructor
+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/2025.11.03/22.51.21_train_llm_lowdim_parking-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-parking-v0/checkpoint-5700
+  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}

2025.12.17/04.16.11_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+hydra:
+  run:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  sweep:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - policy.loss_llm_weight=1.0e-3
+  job:
+    name: train
+    chdir: null
+    override_dirname: policy.loss_llm_weight=1.0e-3
+    id: ???
+    num: ???
+    config_name: llmdp_parking-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: /root/workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /root/workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /root/workspace/LLM-BC/data/outputs/2025.12.17/04.16.11_train_llm_diffusion_unet_lowdim_parking-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

2025.12.17/04.16.11_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- policy.loss_llm_weight=1.0e-3

2025.12.17/04.16.11_train_llm_diffusion_unet_lowdim_parking-v0/train.log

@@ -0,0 +1,7 @@
+[2025-12-17 04:16:12,493][numexpr.utils][INFO] - Note: detected 128 virtual cores but NumExpr set to maximum of 64, check "NUMEXPR_MAX_THREADS" environment variable.
+[2025-12-17 04:16:12,493][numexpr.utils][INFO] - Note: NumExpr detected 128 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 16.
+[2025-12-17 04:16:12,493][numexpr.utils][INFO] - NumExpr defaulting to 16 threads.
+[2025-12-17 04:16:14,423][datasets][INFO] - PyTorch version 2.2.2 available.
+[2025-12-17 04:16:14,424][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2025-12-17 04:16:14,425][datasets][INFO] - JAX version 0.4.30 available.
+[2025-12-17 04:16:15,852][llmbc.model.diffusion.conditional_unet1d][INFO] - number of parameters: 6.514023e+07

2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-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: 0.001
+  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: 42
+  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: parking-v0
+  obs_dim: 18
+  action_dim: 2
+  env_runner:
+    _target_: llmbc.env_runner.highway_lowdim_runner.HighwayLowdimRunner
+    env_name: llf-highway-parking-v0
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 80
+    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.highway_lowdim_dataset.HighwayLowdimDataset
+    data_path: datasets/parking-v0.pt
+    data_path2: datasets/parking-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.highway_instructor.parking_v0_instructor.ParkingV0Instructor
+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/2025.11.03/22.51.21_train_llm_lowdim_parking-v0/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-parking-v0/checkpoint-5700
+  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}

2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/hydra.yaml

@@ -0,0 +1,155 @@
+hydra:
+  run:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  sweep:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - policy.loss_llm_weight=1.0e-3
+  job:
+    name: train
+    chdir: null
+    override_dirname: policy.loss_llm_weight=1.0e-3
+    id: ???
+    num: ???
+    config_name: llmdp_parking-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: /root/workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /root/workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /root/workspace/LLM-BC/data/outputs/2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-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

2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-v0/.hydra/overrides.yaml

@@ -0,0 +1 @@
+- policy.loss_llm_weight=1.0e-3

2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-v0/checkpoints/epoch=0240-test_success_rate=0.960.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3796ac0e06aa2ff3ab48a9b1c5603d32060354b3f97aa848effac529ec84612f
+size 1042506738

2025.12.17/04.26.39_train_llm_diffusion_unet_lowdim_parking-v0/checkpoints/epoch=0335-test_success_rate=0.940.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ea234931279169c34ad5d572a37f80ac85c64efb2762f4484c42888a8394bc2
+size 1042506802