Upload folder using huggingface_hub

bimanual_rope_cam0/.hydra/config.yaml

@@ -0,0 +1,291 @@
+experiment:
+  debug: ${debug}
+  tasks:
+  - training
+  num_nodes: 1
+  num_devices: 1
+  training:
+    precision: 32-true
+    compile: false
+    lr: 8.0e-05
+    batch_size: 4
+    max_epochs: -1
+    max_steps: 1000005
+    max_time: null
+    data:
+      num_workers: 4
+      shuffle: true
+    optim:
+      accumulate_grad_batches: 1
+      gradient_clip_val: 1.0
+    checkpointing:
+      every_n_train_steps: 10000
+      every_n_epochs: null
+      train_time_interval: null
+      enable_version_counter: false
+    log_every_n_steps: 100
+  validation:
+    precision: 16-mixed
+    compile: false
+    batch_size: 2
+    val_every_n_step: 30000
+    val_every_n_epoch: null
+    limit_batch: 1.0
+    inference_mode: true
+    data:
+      num_workers: 4
+      shuffle: false
+  test:
+    precision: 16-mixed
+    compile: false
+    batch_size: 8
+    limit_batch: 1
+    data:
+      num_workers: 16
+      shuffle: false
+  logging:
+    metrics:
+    - fvd
+dataset:
+  debug: ${debug}
+  dataset_dir: /work/hdd/bcyd/ywang41/diffusion-forcing/data/real_aloha/bimanual_rope_1201
+  horizon: 10
+  aug_mode: none
+  n_frames: ${dataset.horizon}
+  observation_shape:
+  - 3
+  - ${dataset.resolution}
+  - ${dataset.resolution}
+  context_length: 2
+  frame_skip: ${dataset.skip_frame}
+  skip_frame: 1
+  pad_after: 7
+  pad_before: 1
+  rotation_rep: rotation_6d
+  seed: 42
+  val_ratio: 0.1
+  skip_idx: 1
+  use_cache: true
+  resolution: 128
+  obs_keys:
+  - camera_0_color
+  low_dim_keys: []
+  delta_action: false
+  goal_sample: intermediate
+  action_mode: bimanual_rope
+  load_mask: false
+  shape_meta:
+    action:
+      shape:
+      - 4
+    obs:
+      camera_0_color:
+        shape:
+        - 3
+        - ${dataset.resolution}
+        - ${dataset.resolution}
+        type: rgb
+      camera_1_color:
+        shape:
+        - 3
+        - ${dataset.resolution}
+        - ${dataset.resolution}
+        type: rgb
+  val_horizon: 200
+algorithm:
+  debug: ${debug}
+  lr: ${experiment.training.lr}
+  weight_decay: 0.0001
+  warmup_steps: 10000
+  lr_scheduler: linear
+  optimizer_beta:
+  - 0.9
+  - 0.999
+  latent_dim: 512
+  action_dim: 8
+  enc_dim: 64
+  num_components: 1
+  obs_keys: ${dataset.obs_keys}
+  x_shape:
+  - ${eval:'3 * len(${dataset.obs_keys})'}
+  - ${dataset.resolution}
+  - ${dataset.resolution}
+  norm_scale: 6.0
+  num_latent_downsample: 2
+  num_views: ${eval:'len(${dataset.obs_keys})'}
+  num_latent_channel: ${eval:'4 * ${algorithm.num_views}'}
+  latent_resolution: ${eval:'${dataset.resolution} // int(2 ** ${algorithm.num_latent_downsample})'}
+  training_stage: 2
+  load_ae: ${eval:'None'}
+  dtype: ${torch:float}
+  mask_prev_action: false
+  device: cuda
+  noise_level: log_normal
+  val_render: true
+  scheduling_matrix: autoregressive
+  uncertainty_scale: 1.0
+  guidance_scale: 1.0
+  n_frames: ${dataset.horizon}
+  dyn_infer_steps: 1
+  dec_infer_steps: 3
+  last_frame_loss_only: false
+  prev_frame_noise_scale: 0.1
+  robust_latent: false
+  delta: ${eval:'0.00054 * ${algorithm.num_latent_channel} * ${algorithm.latent_resolution}
+    * ${algorithm.latent_resolution}'}
+  sampling_strategy: terminal_only
+  sampling_strategy_params: []
+  dynamics:
+    _target_: interactive_world_sim.algorithms.latent_dynamics.models.cm_latent_dynamics.CMLatentDynamics
+    action_dim: ${algorithm.action_dim}
+    latent_dim: ${algorithm.num_latent_channel}
+    dim: 64
+    action_emb_dim: 512
+    resnet_block_groups: 8
+    dim_mults:
+    - 1
+    - 2
+    attn_dim_head: 128
+    attn_heads: 4
+    use_linear_attn: true
+    use_init_temporal_attn: true
+    init_kernel_size: 5
+    is_causal: true
+    time_emb_type: rotary
+    dtype: ${algorithm.dtype}
+  noise_scheduler:
+    _target_: interactive_world_sim.utils.cm_utils.DDPMScheduler
+    x_shape: ${algorithm.x_shape}
+    timesteps: ${algorithm.diffusion.timesteps}
+    sampling_timesteps: ${algorithm.diffusion.sampling_timesteps}
+    beta_schedule: ${algorithm.diffusion.beta_schedule}
+    schedule_fn_kwargs: ${algorithm.diffusion.schedule_fn_kwargs}
+    objective: ${algorithm.diffusion.objective}
+    loss_weighting: uniform
+    snr_clip: 5.0
+    cum_snr_decay: ${algorithm.diffusion.cum_snr_decay}
+    ddim_sampling_eta: ${algorithm.diffusion.ddim_sampling_eta}
+    clip_noise: ${algorithm.diffusion.clip_noise}
+    stabilization_level: ${algorithm.diffusion.stabilization_level}
+    dtype: ${algorithm.dtype}
+  diffusion:
+    beta_schedule: sigmoid
+    objective: pred_v
+    use_fused_snr: true
+    cum_snr_decay: 0.96
+    clip_noise: 6.0
+    schedule_fn_kwargs: {}
+    timesteps: 1000
+    sampling_timesteps: 50
+    ddim_sampling_eta: 0.0
+    snr_clip: 5.0
+    model_channels: ${algorithm.enc_dim}
+    num_latent_downsample: ${algorithm.num_latent_downsample}
+    num_latent_channel: ${algorithm.num_latent_channel}
+    num_res_blocks: 2
+    attention_resolutions:
+    - 2
+    - 4
+    - 8
+    dropout: 0.1
+    channel_mult:
+    - 1
+    - 2
+    - 3
+    num_head_channels: 64
+    resblock_updown: true
+    use_scale_shift_norm: true
+    num_components: ${algorithm.num_components}
+    image_size: ${dataset.resolution}
+    stabilization_level: 15
+  metrics:
+  - fvd
+cost_fn:
+  debug: ${debug}
+  lr: ${experiment.training.lr}
+  cfg_ae:
+    debug: ${debug}
+    lr: ${experiment.training.lr}
+    weight_decay: 0.0001
+    warmup_steps: 10000
+    optimizer_beta:
+    - 0.9
+    - 0.999
+    latent_dim: 32
+    action_dim: 10
+    enc_dim: 64
+    num_components: 1
+    obs_keys: ${dataset.obs_keys}
+    x_shape:
+    - ${eval:'3 * len(${dataset.obs_keys})'}
+    - ${dataset.resolution}
+    - ${dataset.resolution}
+    norm_scale: 6.0
+    num_latent_downsample: 2
+    num_latent_channel: 4
+    training_stage: 1
+    load_ae: ${eval:'None'}
+    device: cuda
+    noise_level: random_all
+    dynamics:
+      _target_: algorithms.latent_dynamics.models.film_conv2d_dyn_v3.FiLMConv2DV3Dyn
+      horizon: 9
+      action_dim: ${cost_fn.cfg_ae.action_dim}
+      cond_dim: ${cost_fn.cfg_ae.num_latent_channel}
+    diffusion:
+      beta_schedule: sigmoid
+      objective: pred_v
+      use_fused_snr: true
+      cum_snr_decay: 0.96
+      clip_noise: 6.0
+      schedule_fn_kwargs: {}
+      timesteps: 1000
+      sampling_timesteps: 50
+      ddim_sampling_eta: 0.0
+      snr_clip: 5.0
+      model_channels: ${cost_fn.cfg_ae.enc_dim}
+      num_latent_downsample: ${cost_fn.cfg_ae.num_latent_downsample}
+      num_latent_channel: ${cost_fn.cfg_ae.num_latent_channel}
+      num_res_blocks: 2
+      attention_resolutions:
+      - 2
+      - 4
+      - 8
+      dropout: 0.1
+      channel_mult:
+      - 1
+      - 2
+      - 3
+      num_head_channels: 64
+      resblock_updown: true
+      use_scale_shift_norm: true
+      num_components: ${cost_fn.cfg_ae.num_components}
+      image_size: ${dataset.resolution}
+      stabilization_level: 15
+    metrics:
+    - fvd
+  weight_decay: 0.0001
+  warmup_steps: 10000
+  optimizer_beta:
+  - 0.9
+  - 0.999
+  obs_keys: ${dataset.obs_keys}
+  x_shape:
+  - ${eval:'${cost_fn.cfg_ae.num_latent_channel} * len(${dataset.obs_keys})'}
+  - ${eval:'${dataset.resolution} // int(2 ** ${cost_fn.cfg_ae.num_latent_downsample})'}
+  - ${eval:'${dataset.resolution} // int(2 ** ${cost_fn.cfg_ae.num_latent_downsample})'}
+  load_ae: ${eval:'None'}
+  latent_dim: 512
+  enc_dim: 16
+  mlp_dims:
+  - 1024
+  - 512
+  device: cuda
+debug: false
+wandb:
+  entity: yixuan1999
+  project: diffusion-forcing
+  mode: online
+resume: null
+load: null
+name: train_dyn_latent_student_v1_4_bimanual_rope_1201_cam_0_stage_2

bimanual_rope_cam0/.hydra/hydra.yaml

@@ -0,0 +1,193 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    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:
+    - +name=train_dyn_latent_student_v1_4_bimanual_rope_1201_cam_0_stage_2
+    - algorithm=student_v1_4
+    - experiment=exp_latent_dyn
+    - dataset=real_aloha_dataset
+    - dataset.dataset_dir=/work/hdd/bcyd/ywang41/diffusion-forcing/data/real_aloha/bimanual_rope_1201
+    - dataset.horizon=10
+    - +dataset.val_horizon=200
+    - dataset.skip_frame=1
+    - dataset.obs_keys=[camera_0_color]
+    - dataset.action_mode=bimanual_rope
+    - experiment.training.batch_size=4
+    - experiment.training.max_steps=1000005
+    - experiment.training.log_every_n_steps=100
+    - experiment.validation.limit_batch=1.0
+    - experiment.validation.batch_size=2
+    - experiment.validation.val_every_n_step=30000
+    - experiment.training.checkpointing.every_n_train_steps=10000
+    - experiment.training.data.num_workers=4
+    - experiment.validation.data.num_workers=4
+    - algorithm.latent_dim=512
+    - algorithm.action_dim=8
+    - algorithm.training_stage=2
+    - algorithm.mask_prev_action=False
+    - algorithm.lr_scheduler=linear
+    - algorithm.noise_scheduler.loss_weighting=uniform
+    - algorithm.noise_scheduler.snr_clip=5.0
+    - algorithm.sampling_strategy=terminal_only
+    - algorithm.sampling_strategy_params=[]
+    - algorithm.dyn_infer_steps=1
+    - algorithm.dec_infer_steps=3
+    - algorithm.prev_frame_noise_scale=0.1
+    - algorithm.last_frame_loss_only=False
+    - algorithm.load_ae=outputs/2025-12-07/13-24-13/checkpoints/epoch\=29-step\=880000.ckpt
+  job:
+    name: main
+    chdir: null
+    override_dirname: +dataset.val_horizon=200,+name=train_dyn_latent_student_v1_4_bimanual_rope_1201_cam_0_stage_2,algorithm.action_dim=8,algorithm.dec_infer_steps=3,algorithm.dyn_infer_steps=1,algorithm.last_frame_loss_only=False,algorithm.latent_dim=512,algorithm.load_ae=outputs/2025-12-07/13-24-13/checkpoints/epoch\=29-step\=880000.ckpt,algorithm.lr_scheduler=linear,algorithm.mask_prev_action=False,algorithm.noise_scheduler.loss_weighting=uniform,algorithm.noise_scheduler.snr_clip=5.0,algorithm.prev_frame_noise_scale=0.1,algorithm.sampling_strategy=terminal_only,algorithm.sampling_strategy_params=[],algorithm.training_stage=2,algorithm=student_v1_4,dataset.action_mode=bimanual_rope,dataset.dataset_dir=/work/hdd/bcyd/ywang41/diffusion-forcing/data/real_aloha/bimanual_rope_1201,dataset.horizon=10,dataset.obs_keys=[camera_0_color],dataset.skip_frame=1,dataset=real_aloha_dataset,experiment.training.batch_size=4,experiment.training.checkpointing.every_n_train_steps=10000,experiment.training.data.num_workers=4,experiment.training.log_every_n_steps=100,experiment.training.max_steps=1000005,experiment.validation.batch_size=2,experiment.validation.data.num_workers=4,experiment.validation.limit_batch=1.0,experiment.validation.val_every_n_step=30000,experiment=exp_latent_dyn
+    id: ???
+    num: ???
+    config_name: config
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /projects/bcyd/ywang41/diffusion-forcing
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /projects/bcyd/ywang41/diffusion-forcing/configurations
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /projects/bcyd/ywang41/diffusion-forcing/outputs/2026-01-07/23-27-27
+    choices:
+      cost_fn: cost_fn_v0_2
+      cost_fn/cfg_ae: latent_dyn_v3_2_film
+      cluster: null
+      algorithm: student_v1_4
+      dataset: real_aloha_dataset
+      experiment: exp_latent_dyn
+      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

bimanual_rope_cam0/.hydra/overrides.yaml

@@ -0,0 +1,33 @@
+- +name=train_dyn_latent_student_v1_4_bimanual_rope_1201_cam_0_stage_2
+- algorithm=student_v1_4
+- experiment=exp_latent_dyn
+- dataset=real_aloha_dataset
+- dataset.dataset_dir=/work/hdd/bcyd/ywang41/diffusion-forcing/data/real_aloha/bimanual_rope_1201
+- dataset.horizon=10
+- +dataset.val_horizon=200
+- dataset.skip_frame=1
+- dataset.obs_keys=[camera_0_color]
+- dataset.action_mode=bimanual_rope
+- experiment.training.batch_size=4
+- experiment.training.max_steps=1000005
+- experiment.training.log_every_n_steps=100
+- experiment.validation.limit_batch=1.0
+- experiment.validation.batch_size=2
+- experiment.validation.val_every_n_step=30000
+- experiment.training.checkpointing.every_n_train_steps=10000
+- experiment.training.data.num_workers=4
+- experiment.validation.data.num_workers=4
+- algorithm.latent_dim=512
+- algorithm.action_dim=8
+- algorithm.training_stage=2
+- algorithm.mask_prev_action=False
+- algorithm.lr_scheduler=linear
+- algorithm.noise_scheduler.loss_weighting=uniform
+- algorithm.noise_scheduler.snr_clip=5.0
+- algorithm.sampling_strategy=terminal_only
+- algorithm.sampling_strategy_params=[]
+- algorithm.dyn_infer_steps=1
+- algorithm.dec_infer_steps=3
+- algorithm.prev_frame_noise_scale=0.1
+- algorithm.last_frame_loss_only=False
+- algorithm.load_ae=outputs/2025-12-07/13-24-13/checkpoints/epoch\=29-step\=880000.ckpt

bimanual_rope_cam0/checkpoints/best.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fd50e6b05eed1866ded620fe5177dc00fed0a6d9d8cf969f72b0651382944a1
+size 232147820