Upload policy weights, train config and readme

README.md

@@ -5,9 +5,9 @@ license: apache-2.0
 model_name: act
 pipeline_tag: robotics
 tags:
-- act
-- lerobot
 - robotics
+- lerobot
+- act
 ---
 
 # Model Card for act

config.json

@@ -42,8 +42,8 @@
     "tags": null,
     "license": null,
     "pretrained_path": null,
-    "chunk_size": 100,
-    "n_action_steps": 100,
+    "chunk_size": 50,
+    "n_action_steps": 50,
     "normalization_mapping": {
         "VISUAL": "MEAN_STD",
         "STATE": "MEAN_STD",
@@ -51,21 +51,21 @@
     },
     "vision_backbone": "resnet18",
     "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
-    "replace_final_stride_with_dilation": false,
+    "replace_final_stride_with_dilation": 0,
     "pre_norm": false,
-    "dim_model": 512,
-    "n_heads": 8,
-    "dim_feedforward": 3200,
+    "dim_model": 256,
+    "n_heads": 4,
+    "dim_feedforward": 800,
     "feedforward_activation": "relu",
-    "n_encoder_layers": 4,
+    "n_encoder_layers": 2,
     "n_decoder_layers": 1,
     "use_vae": true,
-    "latent_dim": 32,
+    "latent_dim": 16,
     "n_vae_encoder_layers": 4,
     "temporal_ensemble_coeff": null,
-    "dropout": 0.1,
+    "dropout": 0.3,
     "kl_weight": 10.0,
-    "optimizer_lr": 1e-05,
+    "optimizer_lr": 0.0001,
     "optimizer_weight_decay": 0.0001,
     "optimizer_lr_backbone": 1e-05
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5c027bf3cf460ae8b812eea0015e049d48e5409712fe8d3d3f8adcf9ec0790e
-size 206716132
+oid sha256:ab8f85eccbbabe8d3939009d3c565d2b1521e47fad0144f41c1791230eaf944a
+size 65438420

train_config.json

@@ -124,8 +124,8 @@
         "tags": null,
         "license": null,
         "pretrained_path": null,
-        "chunk_size": 100,
-        "n_action_steps": 100,
+        "chunk_size": 50,
+        "n_action_steps": 50,
         "normalization_mapping": {
             "VISUAL": "MEAN_STD",
             "STATE": "MEAN_STD",
@@ -133,21 +133,21 @@
         },
         "vision_backbone": "resnet18",
         "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
-        "replace_final_stride_with_dilation": false,
+        "replace_final_stride_with_dilation": 0,
         "pre_norm": false,
-        "dim_model": 512,
-        "n_heads": 8,
-        "dim_feedforward": 3200,
+        "dim_model": 256,
+        "n_heads": 4,
+        "dim_feedforward": 800,
         "feedforward_activation": "relu",
-        "n_encoder_layers": 4,
+        "n_encoder_layers": 2,
         "n_decoder_layers": 1,
         "use_vae": true,
-        "latent_dim": 32,
+        "latent_dim": 16,
         "n_vae_encoder_layers": 4,
         "temporal_ensemble_coeff": null,
-        "dropout": 0.1,
+        "dropout": 0.3,
         "kl_weight": 10.0,
-        "optimizer_lr": 1e-05,
+        "optimizer_lr": 0.0001,
         "optimizer_weight_decay": 0.0001,
         "optimizer_lr_backbone": 1e-05
     },
@@ -157,16 +157,16 @@
     "seed": 1000,
     "num_workers": 4,
     "batch_size": 24,
-    "steps": 100,
-    "eval_freq": 20000,
-    "log_freq": 200,
+    "steps": 500,
+    "eval_freq": 200,
+    "log_freq": 50,
     "tolerance_s": 0.0001,
     "save_checkpoint": true,
-    "save_freq": 20000,
+    "save_freq": 200,
     "use_policy_training_preset": true,
     "optimizer": {
         "type": "adamw",
-        "lr": 1e-05,
+        "lr": 0.0001,
         "weight_decay": 0.0001,
         "grad_clip_norm": 10.0,
         "betas": [