Upload 663 files

Upload the visualization folder, which contains all the plots that present in the paper, as well as all the training records in .csv files that generates these plots.
You can reproduce all the figures using this folder, following the instructions in the docs/ReproduceFigs.md document.

visualize/0_Models/DiffusionModel/can/25-04-16-20-33-20/cfg.yaml

@@ -0,0 +1,111 @@
+_target_: agent.eval.eval_diffusion_img_agent.EvalImgDiffusionAgent
+name: can_eval_diffusion_mlp_img_ta4_td100
+logdir: ${oc.env:REINFLOW_LOG_DIR}/robomimic/eval/can_eval_diffusion_mlp_img_ta4_td100/2025-04-16_20-32-19_42
+base_policy_path: log/robomimic/pretrain/can/DDPM/state_5000.pt
+robomimic_env_cfg_path: cfg/robomimic/env_meta/can-img.json
+normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/can-img/normalization.npz
+seed: 42
+device: cuda:4
+env_name: can
+obs_dim: 9
+action_dim: 7
+denoising_steps: 100
+cond_steps: 1
+img_cond_steps: 1
+horizon_steps: 4
+act_steps: 4
+use_ddim: true
+ddim_steps: 5
+n_steps: 300
+render_num: 0
+env:
+  n_envs: 50
+  name: can
+  best_reward_threshold_for_success: 1
+  max_episode_steps: 300
+  save_video: false
+  use_image_obs: true
+  wrappers:
+    robomimic_image:
+      normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/can-img/normalization.npz
+      low_dim_keys:
+      - robot0_eef_pos
+      - robot0_eef_quat
+      - robot0_gripper_qpos
+      image_keys:
+      - robot0_eye_in_hand_image
+      shape_meta:
+        obs:
+          rgb:
+            shape:
+            - 3
+            - 96
+            - 96
+          state:
+            shape:
+            - 9
+        action:
+          shape:
+          - 7
+    multi_step:
+      n_obs_steps: 1
+      n_action_steps: 4
+      max_episode_steps: 300
+      reset_within_step: true
+  render: true
+shape_meta:
+  obs:
+    rgb:
+      shape:
+      - 3
+      - 96
+      - 96
+    state:
+      shape:
+      - 9
+  action:
+    shape:
+    - 7
+model:
+  _target_: model.diffusion.diffusion.DiffusionModel
+  predict_epsilon: true
+  denoised_clip_value: 1.0
+  randn_clip_value: 3
+  use_ddim: true
+  ddim_steps: 5
+  network_path: log/robomimic/pretrain/can/DDPM/state_5000.pt
+  network:
+    _target_: model.diffusion.mlp_diffusion.VisionDiffusionMLP
+    backbone:
+      _target_: model.common.vit.VitEncoder
+      obs_shape:
+      - 3
+      - 96
+      - 96
+      num_channel: 3
+      img_h: 96
+      img_w: 96
+      cfg:
+        patch_size: 8
+        depth: 1
+        embed_dim: 128
+        num_heads: 4
+        embed_style: embed2
+        embed_norm: 0
+    augment: false
+    spatial_emb: 128
+    time_dim: 32
+    mlp_dims:
+    - 512
+    - 512
+    - 512
+    residual_style: true
+    img_cond_steps: 1
+    cond_dim: 9
+    horizon_steps: 4
+    action_dim: 7
+  horizon_steps: 4
+  obs_dim: 9
+  action_dim: 7
+  denoising_steps: 100
+  device: cuda:4

visualize/0_Models/DiffusionModel/can/25-04-16-20-33-20/eval_statistics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ac3763debf5cae78ba90db3f2c936d16110b57fc4cc69cf48e2dd2e1de1d6a9
+size 1270

visualize/0_Models/DiffusionModel/can/25-04-17-11-13-29/cfg.yaml

@@ -0,0 +1,111 @@
+_target_: agent.eval.eval_diffusion_img_agent.EvalImgDiffusionAgent
+name: can_eval_diffusion_mlp_img_ta4_td100
+logdir: ${oc.env:REINFLOW_LOG_DIR}/robomimic/eval/can_eval_diffusion_mlp_img_ta4_td100/2025-04-17_11-12-22_42
+base_policy_path: log/robomimic/pretrain/can/DDPM/state_5000.pt
+robomimic_env_cfg_path: cfg/robomimic/env_meta/can-img.json
+normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/can-img/normalization.npz
+seed: 42
+device: cuda:2
+env_name: can
+obs_dim: 9
+action_dim: 7
+denoising_steps: 100
+cond_steps: 1
+img_cond_steps: 1
+horizon_steps: 4
+act_steps: 4
+use_ddim: true
+ddim_steps: 5
+n_steps: 300
+render_num: 0
+env:
+  n_envs: 50
+  name: can
+  best_reward_threshold_for_success: 1
+  max_episode_steps: 300
+  save_video: false
+  use_image_obs: true
+  wrappers:
+    robomimic_image:
+      normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/can-img/normalization.npz
+      low_dim_keys:
+      - robot0_eef_pos
+      - robot0_eef_quat
+      - robot0_gripper_qpos
+      image_keys:
+      - robot0_eye_in_hand_image
+      shape_meta:
+        obs:
+          rgb:
+            shape:
+            - 3
+            - 96
+            - 96
+          state:
+            shape:
+            - 9
+        action:
+          shape:
+          - 7
+    multi_step:
+      n_obs_steps: 1
+      n_action_steps: 4
+      max_episode_steps: 300
+      reset_within_step: true
+  render: true
+shape_meta:
+  obs:
+    rgb:
+      shape:
+      - 3
+      - 96
+      - 96
+    state:
+      shape:
+      - 9
+  action:
+    shape:
+    - 7
+model:
+  _target_: model.diffusion.diffusion.DiffusionModel
+  predict_epsilon: true
+  denoised_clip_value: 1.0
+  randn_clip_value: 3
+  use_ddim: true
+  ddim_steps: 5
+  network_path: log/robomimic/pretrain/can/DDPM/state_5000.pt
+  network:
+    _target_: model.diffusion.mlp_diffusion.VisionDiffusionMLP
+    backbone:
+      _target_: model.common.vit.VitEncoder
+      obs_shape:
+      - 3
+      - 96
+      - 96
+      num_channel: 3
+      img_h: 96
+      img_w: 96
+      cfg:
+        patch_size: 8
+        depth: 1
+        embed_dim: 128
+        num_heads: 4
+        embed_style: embed2
+        embed_norm: 0
+    augment: false
+    spatial_emb: 128
+    time_dim: 32
+    mlp_dims:
+    - 512
+    - 512
+    - 512
+    residual_style: true
+    img_cond_steps: 1
+    cond_dim: 9
+    horizon_steps: 4
+    action_dim: 7
+  horizon_steps: 4
+  obs_dim: 9
+  action_dim: 7
+  denoising_steps: 100
+  device: cuda:2

visualize/0_Models/DiffusionModel/can/25-04-17-11-13-29/eval_statistics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:807ac2188fd6909b7fbd91d2a13bebe3f4e28d52a05706b6b85d56248b34117f
+size 2838

visualize/0_Models/DiffusionModel/square/25-04-16-20-33-18/cfg.yaml

@@ -0,0 +1,110 @@
+_target_: agent.eval.eval_diffusion_img_agent.EvalImgDiffusionAgent
+name: square_eval_diffusion_mlp_img_ta4_td100
+logdir: ${oc.env:REINFLOW_LOG_DIR}/robomimic/eval/square_eval_diffusion_mlp_img_ta4_td100/2025-04-16_20-32-23_42
+base_policy_path: log/robomimic/pretrain/square/DDPM/state_4000.pt
+robomimic_env_cfg_path: cfg/robomimic/env_meta/square-img.json
+normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/square-img/normalization.npz
+seed: 42
+device: cuda:3
+env_name: square
+obs_dim: 9
+action_dim: 7
+denoising_steps: 100
+cond_steps: 1
+img_cond_steps: 1
+horizon_steps: 4
+act_steps: 4
+use_ddim: true
+ddim_steps: 5
+n_steps: 300
+render_num: 0
+env:
+  n_envs: 50
+  name: square
+  best_reward_threshold_for_success: 1
+  max_episode_steps: 300
+  save_video: false
+  use_image_obs: true
+  wrappers:
+    robomimic_image:
+      normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/square-img/normalization.npz
+      low_dim_keys:
+      - robot0_eef_pos
+      - robot0_eef_quat
+      - robot0_gripper_qpos
+      image_keys:
+      - agentview_image
+      shape_meta:
+        obs:
+          rgb:
+            shape:
+            - 3
+            - 96
+            - 96
+          state:
+            shape:
+            - 9
+        action:
+          shape:
+          - 7
+    multi_step:
+      n_obs_steps: 1
+      n_action_steps: 4
+      max_episode_steps: 300
+      reset_within_step: true
+  render: true
+shape_meta:
+  obs:
+    rgb:
+      shape:
+      - 3
+      - 96
+      - 96
+    state:
+      shape:
+      - 9
+  action:
+    shape:
+    - 7
+model:
+  _target_: model.diffusion.diffusion.DiffusionModel
+  predict_epsilon: true
+  denoised_clip_value: 1.0
+  randn_clip_value: 3
+  use_ddim: true
+  ddim_steps: 5
+  network_path: log/robomimic/pretrain/square/DDPM/state_4000.pt
+  network:
+    _target_: model.diffusion.mlp_diffusion.VisionDiffusionMLP
+    backbone:
+      _target_: model.common.vit.VitEncoder
+      obs_shape:
+      - 3
+      - 96
+      - 96
+      num_channel: 3
+      cfg:
+        patch_size: 8
+        depth: 1
+        embed_dim: 128
+        num_heads: 4
+        embed_style: embed2
+        embed_norm: 0
+    img_cond_steps: 1
+    augment: true
+    num_img: 1
+    spatial_emb: 128
+    time_dim: 32
+    mlp_dims:
+    - 768
+    - 768
+    - 768
+    residual_style: true
+    cond_dim: 9
+    horizon_steps: 4
+    action_dim: 7
+  horizon_steps: 4
+  obs_dim: 9
+  action_dim: 7
+  denoising_steps: 100
+  device: cuda:3

visualize/0_Models/DiffusionModel/square/25-04-16-20-33-18/eval_statistics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13e75203906317320aae924f2b31e0c43b7a38939dcd18bc6904f1ec73dcc22e
+size 1270

visualize/0_Models/DiffusionModel/square/25-04-17-11-13-30/cfg.yaml

@@ -0,0 +1,110 @@
+_target_: agent.eval.eval_diffusion_img_agent.EvalImgDiffusionAgent
+name: square_eval_diffusion_mlp_img_ta4_td100
+logdir: ${oc.env:REINFLOW_LOG_DIR}/robomimic/eval/square_eval_diffusion_mlp_img_ta4_td100/2025-04-17_11-12-29_42
+base_policy_path: log/robomimic/pretrain/square/DDPM/state_4000.pt
+robomimic_env_cfg_path: cfg/robomimic/env_meta/square-img.json
+normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/square-img/normalization.npz
+seed: 42
+device: cuda:3
+env_name: square
+obs_dim: 9
+action_dim: 7
+denoising_steps: 100
+cond_steps: 1
+img_cond_steps: 1
+horizon_steps: 4
+act_steps: 4
+use_ddim: true
+ddim_steps: 5
+n_steps: 300
+render_num: 0
+env:
+  n_envs: 50
+  name: square
+  best_reward_threshold_for_success: 1
+  max_episode_steps: 300
+  save_video: false
+  use_image_obs: true
+  wrappers:
+    robomimic_image:
+      normalization_path: ${oc.env:REINFLOW_DATA_DIR}/robomimic/square-img/normalization.npz
+      low_dim_keys:
+      - robot0_eef_pos
+      - robot0_eef_quat
+      - robot0_gripper_qpos
+      image_keys:
+      - agentview_image
+      shape_meta:
+        obs:
+          rgb:
+            shape:
+            - 3
+            - 96
+            - 96
+          state:
+            shape:
+            - 9
+        action:
+          shape:
+          - 7
+    multi_step:
+      n_obs_steps: 1
+      n_action_steps: 4
+      max_episode_steps: 300
+      reset_within_step: true
+  render: true
+shape_meta:
+  obs:
+    rgb:
+      shape:
+      - 3
+      - 96
+      - 96
+    state:
+      shape:
+      - 9
+  action:
+    shape:
+    - 7
+model:
+  _target_: model.diffusion.diffusion.DiffusionModel
+  predict_epsilon: true
+  denoised_clip_value: 1.0
+  randn_clip_value: 3
+  use_ddim: true
+  ddim_steps: 5
+  network_path: log/robomimic/pretrain/square/DDPM/state_4000.pt
+  network:
+    _target_: model.diffusion.mlp_diffusion.VisionDiffusionMLP
+    backbone:
+      _target_: model.common.vit.VitEncoder
+      obs_shape:
+      - 3
+      - 96
+      - 96
+      num_channel: 3
+      cfg:
+        patch_size: 8
+        depth: 1
+        embed_dim: 128
+        num_heads: 4
+        embed_style: embed2
+        embed_norm: 0
+    img_cond_steps: 1
+    augment: true
+    num_img: 1
+    spatial_emb: 128
+    time_dim: 32
+    mlp_dims:
+    - 768
+    - 768
+    - 768
+    residual_style: true
+    cond_dim: 9
+    horizon_steps: 4
+    action_dim: 7
+  horizon_steps: 4
+  obs_dim: 9
+  action_dim: 7
+  denoising_steps: 100
+  device: cuda:3

visualize/0_Models/DiffusionModel/square/25-04-17-11-13-30/eval_statistics.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:715c4204607ec97ec2dafb544bdea63f63692694328e76f4ef890cfc1b4caa2f
+size 2838

visualize/0_Models/DiffusionModel/walker2d-medium-v2/25-04-15-21-10-59/cfg.yaml

@@ -0,0 +1,54 @@
+_target_: agent.eval.eval_diffusion_agent.EvalDiffusionAgent
+name: walker2d-medium-v2_eval_diffusion_mlp_ta4_td20
+logdir: ${oc.env:REINFLOW_LOG_DIR}/gym/eval/walker2d-medium-v2_eval_diffusion_mlp_ta4_td20/2025-04-15_21-10-52_42
+base_policy_path: log/gym/pretrain/walker2d-medium-v2_pre_diffusion_mlp_ta4_td20/2024-06-12_23-06-12/checkpoint/state_3000.pt
+normalization_path: ${oc.env:REINFLOW_DATA_DIR}/gym/walker2d-medium-v2/normalization.npz
+seed: 42
+device: cuda:0
+env_name: walker2d-medium-v2
+obs_dim: 17
+action_dim: 6
+denoising_steps: 20
+cond_steps: 1
+horizon_steps: 4
+act_steps: 4
+batch_size: 128
+n_steps: 500
+render_num: 0
+env:
+  n_envs: 40
+  name: walker2d-medium-v2
+  max_episode_steps: 1000
+  reset_at_iteration: false
+  save_video: false
+  best_reward_threshold_for_success: 3
+  wrappers:
+    mujoco_locomotion_lowdim:
+      normalization_path: ${oc.env:REINFLOW_DATA_DIR}/gym/walker2d-medium-v2/normalization.npz
+    multi_step:
+      n_obs_steps: 1
+      n_action_steps: 4
+      max_episode_steps: 1000
+      reset_within_step: true
+model:
+  _target_: model.diffusion.diffusion.DiffusionModel
+  predict_epsilon: true
+  denoised_clip_value: 1.0
+  network_path: log/gym/pretrain/walker2d-medium-v2_pre_diffusion_mlp_ta4_td20/2024-06-12_23-06-12/checkpoint/state_3000.pt
+  network:
+    _target_: model.diffusion.mlp_diffusion.DiffusionMLP
+    time_dim: 16
+    mlp_dims:
+    - 512
+    - 512
+    - 512
+    activation_type: ReLU
+    residual_style: true
+    cond_dim: 17
+    horizon_steps: 4
+    action_dim: 6
+  horizon_steps: 4
+  obs_dim: 17
+  action_dim: 6
+  denoising_steps: 20
+  device: cuda:0

visualize/0_Models/DiffusionModel/walker2d-medium-v2/25-04-15-21-13-28/cfg.yaml

@@ -0,0 +1,54 @@
+_target_: agent.eval.eval_diffusion_agent.EvalDiffusionAgent
+name: walker2d-medium-v2_eval_diffusion_mlp_ta4_td20
+logdir: ${oc.env:REINFLOW_LOG_DIR}/gym/eval/walker2d-medium-v2_eval_diffusion_mlp_ta4_td20/2025-04-15_21-13-20_42
+base_policy_path: log/gym/pretrain/walker2d-medium-v2_pre_diffusion_mlp_ta4_td20/2024-06-12_23-06-12/checkpoint/state_3000.pt
+normalization_path: ${oc.env:REINFLOW_DATA_DIR}/gym/walker2d-medium-v2/normalization.npz
+seed: 42
+device: cuda:0
+env_name: walker2d-medium-v2
+obs_dim: 17
+action_dim: 6
+denoising_steps: 20
+cond_steps: 1
+horizon_steps: 4
+act_steps: 4
+batch_size: 128
+n_steps: 500
+render_num: 0
+env:
+  n_envs: 40
+  name: walker2d-medium-v2
+  max_episode_steps: 1000
+  reset_at_iteration: false
+  save_video: false
+  best_reward_threshold_for_success: 3
+  wrappers:
+    mujoco_locomotion_lowdim:
+      normalization_path: ${oc.env:REINFLOW_DATA_DIR}/gym/walker2d-medium-v2/normalization.npz
+    multi_step:
+      n_obs_steps: 1
+      n_action_steps: 4
+      max_episode_steps: 1000
+      reset_within_step: true
+model:
+  _target_: model.diffusion.diffusion.DiffusionModel
+  predict_epsilon: true
+  denoised_clip_value: 1.0
+  network_path: log/gym/pretrain/walker2d-medium-v2_pre_diffusion_mlp_ta4_td20/2024-06-12_23-06-12/checkpoint/state_3000.pt
+  network:
+    _target_: model.diffusion.mlp_diffusion.DiffusionMLP
+    time_dim: 16
+    mlp_dims:
+    - 512
+    - 512
+    - 512
+    activation_type: ReLU
+    residual_style: true
+    cond_dim: 17
+    horizon_steps: 4
+    action_dim: 6
+  horizon_steps: 4
+  obs_dim: 17
+  action_dim: 6
+  denoising_steps: 20
+  device: cuda:0