Publication of PushT Diffusion Policy model with demonstration videos - 2025-04-28

README.md

@@ -0,0 +1,26 @@
+
+# PushT Diffusion Policy - Robot Control Model
+
+This model is an implementation of Diffusion Policy for the PushT environment, which simulates robotic pushing tasks.
+
+## Model
+
+This model uses a conditional diffusion architecture to predict robotic actions based on visual observations.
+
+## Performance
+
+The model achieves a success rate of 40.0% in the PushT environment with different initial configurations.
+
+## Demonstration Videos
+
+The repository includes demonstration videos in the `videos/` folder.
+
+## Usage
+
+```python
+from lerobot.common.policies.diffusion.modeling_diffusion import DiffusionPolicy
+
+policy = DiffusionPolicy.from_pretrained("RafaelJaime/pusht-diffusion")
+```
+
+Published on 2025-04-28

metadata.json

@@ -0,0 +1,13 @@
+{
+  "library_name": "lerobot",
+  "task_type": "robot-control",
+  "environment": "PushT",
+  "timestamp": "2025-04-28T04:50:36.767363",
+  "videos": [
+    "video_2_experiment_summary.mp4",
+    "video_1_pusht_episode.mp4",
+    "pusht_episode.mp4",
+    "experiment_summary.mp4"
+  ],
+  "device": "cuda"
+}

model/config.json

@@ -0,0 +1,76 @@
+{
+    "type": "diffusion",
+    "n_obs_steps": 2,
+    "normalization_mapping": {
+        "ACTION": "MIN_MAX",
+        "STATE": "MIN_MAX",
+        "VISUAL": "MEAN_STD"
+    },
+    "input_features": {
+        "observation.image": {
+            "type": "VISUAL",
+            "shape": [
+                3,
+                96,
+                96
+            ]
+        },
+        "observation.state": {
+            "type": "STATE",
+            "shape": [
+                2
+            ]
+        }
+    },
+    "output_features": {
+        "action": {
+            "type": "ACTION",
+            "shape": [
+                2
+            ]
+        }
+    },
+    "device": "cuda",
+    "use_amp": false,
+    "horizon": 16,
+    "n_action_steps": 8,
+    "drop_n_last_frames": 7,
+    "vision_backbone": "resnet18",
+    "crop_shape": [
+        84,
+        84
+    ],
+    "crop_is_random": true,
+    "pretrained_backbone_weights": null,
+    "use_group_norm": true,
+    "spatial_softmax_num_keypoints": 32,
+    "use_separate_rgb_encoder_per_camera": false,
+    "down_dims": [
+        512,
+        1024,
+        2048
+    ],
+    "kernel_size": 5,
+    "n_groups": 8,
+    "diffusion_step_embed_dim": 128,
+    "use_film_scale_modulation": true,
+    "noise_scheduler_type": "DDPM",
+    "num_train_timesteps": 100,
+    "beta_schedule": "squaredcos_cap_v2",
+    "beta_start": 0.0001,
+    "beta_end": 0.02,
+    "prediction_type": "epsilon",
+    "clip_sample": true,
+    "clip_sample_range": 1.0,
+    "num_inference_steps": null,
+    "do_mask_loss_for_padding": false,
+    "optimizer_lr": 0.0001,
+    "optimizer_betas": [
+        0.95,
+        0.999
+    ],
+    "optimizer_eps": 1e-08,
+    "optimizer_weight_decay": 1e-06,
+    "scheduler_name": "cosine",
+    "scheduler_warmup_steps": 500
+}

model/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:995d14d35db57d95c35ad9704c3d79c8612b7bc45f3877e5c46c2cdc516856a8
+size 1050862408