Upload policy weights, train config and readme

README.md

@@ -2,20 +2,20 @@
 datasets: lerobot/aloha_sim_transfer_cube_human
 library_name: lerobot
 license: apache-2.0
-model_name: act
+model_name: wam
 pipeline_tag: robotics
 tags:
 - lerobot
-- act
 - robotics
+- wam
 ---
 
-# Model Card for act
+# Model Card for wam
 
 <!-- Provide a quick summary of what the model is/does. -->
 
 
-[Action Chunking with Transformers (ACT)](https://huggingface.co/papers/2304.13705) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates.
+_Model type not recognized — please update this template._
 
 
 This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).

config.json

@@ -1,5 +1,5 @@
 {
-    "type": "act",
+    "type": "wam",
     "n_obs_steps": 1,
     "input_features": {
         "observation.images.top": {
@@ -26,7 +26,7 @@
         }
     },
     "device": "cuda",
-    "use_amp": false,
+    "use_amp": true,
     "use_peft": false,
     "push_to_hub": true,
     "repo_id": "maximellerbach/act_aloha_transfer",
@@ -34,30 +34,41 @@
     "tags": null,
     "license": null,
     "pretrained_path": null,
-    "chunk_size": 100,
-    "n_action_steps": 100,
+    "n_obs_frames": 2,
+    "obs_frame_skip": 5,
+    "pred_frame_skip": 50,
+    "n_pred_frames": 1,
+    "image_size": [
+        256,
+        256
+    ],
+    "horizon": 50,
+    "n_action_steps": 50,
     "normalization_mapping": {
         "VISUAL": "MEAN_STD",
         "STATE": "MEAN_STD",
         "ACTION": "MEAN_STD"
     },
+    "hidden_dim": 384,
+    "dim_feedforward": 1536,
+    "pre_norm": false,
+    "dropout": 0.1,
     "vision_backbone": "resnet18",
     "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,
+    "use_pretrained_backbone": true,
+    "freeze_backbone": false,
     "latent_dim": 32,
     "n_vae_encoder_layers": 4,
-    "temporal_ensemble_coeff": null,
-    "dropout": 0.1,
-    "kl_weight": 10.0,
-    "optimizer_lr": 1e-05,
-    "optimizer_weight_decay": 0.0001,
-    "optimizer_lr_backbone": 1e-05
+    "vae_num_heads": 8,
+    "action_head_dropout": 0.1,
+    "predictor_num_layers": 12,
+    "predictor_num_heads": 12,
+    "predictor_dropout": 0.05,
+    "predictor_max_tokens": 1024,
+    "optimizer_lr": 5e-05,
+    "optimizer_weight_decay": 0.01,
+    "optimizer_grad_clip_norm": 1.0,
+    "scheduler_decay_lr": 1e-05,
+    "scheduler_warmup_steps": 500,
+    "scheduler_decay_steps": 5000
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fa0607c72c92377461b789c3869532b86d4f5b6a44db14c6ded048d75ddf8854
-size 206765304
+oid sha256:87e737f914619f4c71a995e5cae18817eda35b070e563ca423328658272a4fd9
+size 301890384

train_config.json

@@ -120,7 +120,7 @@
         "render_mode": "rgb_array"
     },
     "policy": {
-        "type": "act",
+        "type": "wam",
         "n_obs_steps": 1,
         "input_features": {
             "observation.images.top": {
@@ -147,7 +147,7 @@
             }
         },
         "device": "cuda",
-        "use_amp": false,
+        "use_amp": true,
         "use_peft": false,
         "push_to_hub": true,
         "repo_id": "maximellerbach/act_aloha_transfer",
@@ -155,41 +155,53 @@
         "tags": null,
         "license": null,
         "pretrained_path": null,
-        "chunk_size": 100,
-        "n_action_steps": 100,
+        "n_obs_frames": 2,
+        "obs_frame_skip": 5,
+        "pred_frame_skip": 50,
+        "n_pred_frames": 1,
+        "image_size": [
+            256,
+            256
+        ],
+        "horizon": 50,
+        "n_action_steps": 50,
         "normalization_mapping": {
             "VISUAL": "MEAN_STD",
             "STATE": "MEAN_STD",
             "ACTION": "MEAN_STD"
         },
+        "hidden_dim": 384,
+        "dim_feedforward": 1536,
+        "pre_norm": false,
+        "dropout": 0.1,
         "vision_backbone": "resnet18",
         "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,
+        "use_pretrained_backbone": true,
+        "freeze_backbone": false,
         "latent_dim": 32,
         "n_vae_encoder_layers": 4,
-        "temporal_ensemble_coeff": null,
-        "dropout": 0.1,
-        "kl_weight": 10.0,
-        "optimizer_lr": 1e-05,
-        "optimizer_weight_decay": 0.0001,
-        "optimizer_lr_backbone": 1e-05
+        "vae_num_heads": 8,
+        "action_head_dropout": 0.1,
+        "predictor_num_layers": 12,
+        "predictor_num_heads": 12,
+        "predictor_dropout": 0.05,
+        "predictor_max_tokens": 1024,
+        "optimizer_lr": 5e-05,
+        "optimizer_weight_decay": 0.01,
+        "optimizer_grad_clip_norm": 1.0,
+        "scheduler_decay_lr": 1e-05,
+        "scheduler_warmup_steps": 500,
+        "scheduler_decay_steps": 5000
     },
-    "output_dir": "outputs/train/act_aloha_transfer",
-    "job_name": "aloha_act",
+    "output_dir": "outputs/train/wam_aloha",
+    "job_name": "aloha_wam",
     "resume": false,
     "seed": 1000,
+    "cudnn_deterministic": false,
     "num_workers": 4,
     "batch_size": 8,
-    "steps": 10000,
-    "eval_freq": 20000,
+    "steps": 6000,
+    "eval_freq": 2000,
     "log_freq": 200,
     "tolerance_s": 0.0001,
     "save_checkpoint": true,
@@ -197,29 +209,36 @@
     "use_policy_training_preset": true,
     "optimizer": {
         "type": "adamw",
-        "lr": 1e-05,
-        "weight_decay": 0.0001,
-        "grad_clip_norm": 10.0,
+        "lr": 5e-05,
+        "weight_decay": 0.01,
+        "grad_clip_norm": 1.0,
         "betas": [
             0.9,
             0.999
         ],
         "eps": 1e-08
     },
-    "scheduler": null,
+    "scheduler": {
+        "type": "cosine_decay_with_warmup",
+        "num_warmup_steps": 500,
+        "num_decay_steps": 5000,
+        "peak_lr": 5e-05,
+        "decay_lr": 1e-05
+    },
     "eval": {
-        "n_episodes": 50,
-        "batch_size": 50,
+        "n_episodes": 4,
+        "batch_size": 4,
         "use_async_envs": false
     },
     "wandb": {
-        "enable": false,
+        "enable": true,
         "disable_artifact": false,
         "project": "lerobot",
         "entity": null,
         "notes": null,
-        "run_id": null,
-        "mode": null
+        "run_id": "88n5wijf",
+        "mode": null,
+        "add_tags": true
     },
     "peft": null,
     "use_rabc": false,