Upload folder using huggingface_hub

config.json

@@ -1,24 +1,26 @@
 {
-    "type": "vqbet",
-    "n_obs_steps": 5,
+    "type": "act",
+    "n_obs_steps": 1,
+    "device": "cuda",
+    "use_amp": false,
     "normalization_mapping": {
-        "VISUAL": "IDENTITY",
-        "STATE": "MIN_MAX",
-        "ACTION": "MIN_MAX"
+        "VISUAL": "MEAN_STD",
+        "STATE": "MEAN_STD",
+        "ACTION": "MEAN_STD"
     },
     "input_features": {
-        "observation.image": {
+        "observation.images.top": {
             "type": "VISUAL",
             "shape": [
                 3,
-                96,
-                96
+                480,
+                640
             ]
         },
         "observation.state": {
             "type": "STATE",
             "shape": [
-                2
+                14
             ]
         }
     },
@@ -26,46 +28,29 @@
         "action": {
             "type": "ACTION",
             "shape": [
-                2
+                14
             ]
         }
     },
-    "n_action_pred_token": 3,
-    "action_chunk_size": 5,
+    "chunk_size": 100,
+    "n_action_steps": 100,
     "vision_backbone": "resnet18",
-    "crop_shape": [
-        84,
-        84
-    ],
-    "crop_is_random": true,
-    "pretrained_backbone_weights": null,
-    "use_group_norm": true,
-    "spatial_softmax_num_keypoints": 32,
-    "n_vqvae_training_steps": 20000,
-    "vqvae_n_embed": 16,
-    "vqvae_embedding_dim": 256,
-    "vqvae_enc_hidden_dim": 128,
-    "gpt_block_size": 500,
-    "gpt_input_dim": 512,
-    "gpt_output_dim": 512,
-    "gpt_n_layer": 8,
-    "gpt_n_head": 8,
-    "gpt_hidden_dim": 512,
+    "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
+    "replace_final_stride_with_dilation": false,
+    "pre_norm": false,
+    "dim_model": 512,
+    "n_heads": 8,
+    "dim_feedforward": 3200,
+    "feedforward_activation": "relu",
+    "n_encoder_layers": 4,
+    "n_decoder_layers": 1,
+    "use_vae": true,
+    "latent_dim": 32,
+    "n_vae_encoder_layers": 4,
+    "temporal_ensemble_coeff": null,
     "dropout": 0.1,
-    "mlp_hidden_dim": 1024,
-    "offset_loss_weight": 10000.0,
-    "primary_code_loss_weight": 5.0,
-    "secondary_code_loss_weight": 0.5,
-    "bet_softmax_temperature": 0.1,
-    "sequentially_select": false,
-    "optimizer_lr": 0.0001,
-    "optimizer_betas": [
-        0.95,
-        0.999
-    ],
-    "optimizer_eps": 1e-08,
-    "optimizer_weight_decay": 1e-06,
-    "optimizer_vqvae_lr": 0.001,
-    "optimizer_vqvae_weight_decay": 0.0001,
-    "scheduler_warmup_steps": 500
+    "kl_weight": 10.0,
+    "optimizer_lr": 1e-05,
+    "optimizer_weight_decay": 0.0001,
+    "optimizer_lr_backbone": 1e-05
 }

train_config.json

@@ -1,6 +1,6 @@
 {
     "dataset": {
-        "repo_id": "lerobot/pusht",
+        "repo_id": "lerobot/aloha_sim_insertion_human",
         "episodes": null,
         "image_transforms": {
             "enable": false,
@@ -64,27 +64,27 @@
         "video_backend": "pyav"
     },
     "env": {
-        "type": "pusht",
-        "task": "PushT-v0",
-        "fps": 10,
+        "type": "aloha",
+        "task": "AlohaInsertion-v0",
+        "fps": 50,
         "features": {
             "action": {
                 "type": "ACTION",
                 "shape": [
-                    2
+                    14
                 ]
             },
             "agent_pos": {
                 "type": "STATE",
                 "shape": [
-                    2
+                    14
                 ]
             },
-            "pixels": {
+            "pixels/top": {
                 "type": "VISUAL",
                 "shape": [
-                    384,
-                    384,
+                    480,
+                    640,
                     3
                 ]
             }
@@ -92,36 +92,36 @@
         "features_map": {
             "action": "action",
             "agent_pos": "observation.state",
-            "environment_state": "observation.environment_state",
-            "pixels": "observation.image"
+            "top": "observation.image.top",
+            "pixels/top": "observation.images.top"
         },
-        "episode_length": 300,
+        "episode_length": 400,
         "obs_type": "pixels_agent_pos",
-        "render_mode": "rgb_array",
-        "visualization_width": 384,
-        "visualization_height": 384
+        "render_mode": "rgb_array"
     },
     "policy": {
-        "type": "vqbet",
-        "n_obs_steps": 5,
+        "type": "act",
+        "n_obs_steps": 1,
+        "device": "cuda",
+        "use_amp": false,
         "normalization_mapping": {
-            "VISUAL": "IDENTITY",
-            "STATE": "MIN_MAX",
-            "ACTION": "MIN_MAX"
+            "VISUAL": "MEAN_STD",
+            "STATE": "MEAN_STD",
+            "ACTION": "MEAN_STD"
         },
         "input_features": {
-            "observation.image": {
+            "observation.images.top": {
                 "type": "VISUAL",
                 "shape": [
                     3,
-                    96,
-                    96
+                    480,
+                    640
                 ]
             },
             "observation.state": {
                 "type": "STATE",
                 "shape": [
-                    2
+                    14
                 ]
             }
         },
@@ -129,80 +129,56 @@
             "action": {
                 "type": "ACTION",
                 "shape": [
-                    2
+                    14
                 ]
             }
         },
-        "n_action_pred_token": 3,
-        "action_chunk_size": 5,
+        "chunk_size": 100,
+        "n_action_steps": 100,
         "vision_backbone": "resnet18",
-        "crop_shape": [
-            84,
-            84
-        ],
-        "crop_is_random": true,
-        "pretrained_backbone_weights": null,
-        "use_group_norm": true,
-        "spatial_softmax_num_keypoints": 32,
-        "n_vqvae_training_steps": 20000,
-        "vqvae_n_embed": 16,
-        "vqvae_embedding_dim": 256,
-        "vqvae_enc_hidden_dim": 128,
-        "gpt_block_size": 500,
-        "gpt_input_dim": 512,
-        "gpt_output_dim": 512,
-        "gpt_n_layer": 8,
-        "gpt_n_head": 8,
-        "gpt_hidden_dim": 512,
+        "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
+        "replace_final_stride_with_dilation": false,
+        "pre_norm": false,
+        "dim_model": 512,
+        "n_heads": 8,
+        "dim_feedforward": 3200,
+        "feedforward_activation": "relu",
+        "n_encoder_layers": 4,
+        "n_decoder_layers": 1,
+        "use_vae": true,
+        "latent_dim": 32,
+        "n_vae_encoder_layers": 4,
+        "temporal_ensemble_coeff": null,
         "dropout": 0.1,
-        "mlp_hidden_dim": 1024,
-        "offset_loss_weight": 10000.0,
-        "primary_code_loss_weight": 5.0,
-        "secondary_code_loss_weight": 0.5,
-        "bet_softmax_temperature": 0.1,
-        "sequentially_select": false,
-        "optimizer_lr": 0.0001,
-        "optimizer_betas": [
-            0.95,
-            0.999
-        ],
-        "optimizer_eps": 1e-08,
-        "optimizer_weight_decay": 1e-06,
-        "optimizer_vqvae_lr": 0.001,
-        "optimizer_vqvae_weight_decay": 0.0001,
-        "scheduler_warmup_steps": 500
+        "kl_weight": 10.0,
+        "optimizer_lr": 1e-05,
+        "optimizer_weight_decay": 0.0001,
+        "optimizer_lr_backbone": 1e-05
     },
-    "output_dir": "outputs/train/2025-01-25/20-30-40_pusht_vqbet",
-    "job_name": "pusht_vqbet",
+    "output_dir": "outputs/train/2025-01-25/17-37-25_aloha_act",
+    "job_name": "aloha_act",
     "resume": false,
-    "device": "cuda",
-    "use_amp": false,
-    "seed": 100000,
+    "seed": 1000,
     "num_workers": 4,
-    "batch_size": 64,
-    "steps": 250000,
-    "eval_freq": 25000,
+    "batch_size": 8,
+    "steps": 100000,
+    "eval_freq": 20000,
     "log_freq": 200,
     "save_checkpoint": true,
-    "save_freq": 25000,
+    "save_freq": 20000,
     "use_policy_training_preset": true,
     "optimizer": {
-        "type": "adam",
-        "lr": 0.0001,
+        "type": "adamw",
+        "lr": 1e-05,
         "betas": [
-            0.95,
+            0.9,
             0.999
         ],
         "eps": 1e-08,
-        "weight_decay": 1e-06,
+        "weight_decay": 0.0001,
         "grad_clip_norm": 10.0
     },
-    "scheduler": {
-        "type": "vqbet",
-        "num_warmup_steps": 500,
-        "num_vqvae_training_steps": 20000,
-        "num_cycles": 0.5
-    },
+    "scheduler": null,
     "eval": {
         "n_episodes": 50,
         "batch_size": 50,