Upload policy weights, train config and readme

README.md

@@ -1,6 +1,6 @@
 ---
 base_model: lerobot/smolvla_base
-datasets: Beable/lerobot-SOARM100-sim2
+datasets: Beable/lerobot-SOARM100-sim3
 library_name: lerobot
 license: apache-2.0
 model_name: smolvla

config.json

@@ -51,7 +51,7 @@
     "tokenizer_max_length": 48,
     "num_steps": 10,
     "use_cache": true,
-    "freeze_vision_encoder": true,
+    "freeze_vision_encoder": false,
     "train_expert_only": false,
     "train_state_proj": true,
     "optimizer_lr": 2e-05,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:33b74bfc1e65b4bd1c106aa84ef0b5b422e4a55a4d12287e2d1d2115068759f3
+oid sha256:3e46c3c1de6ef85d549bff9674db9178b62ca1a194a4b6c6830da3f9670d6dfb
 size 1197790016

train_config.json

@@ -1,6 +1,6 @@
 {
     "dataset": {
-        "repo_id": "Beable/lerobot-SOARM100-sim",
+        "repo_id": "Beable/lerobot-SOARM100-sim3",
         "root": null,
         "episodes": null,
         "image_transforms": {
@@ -60,10 +60,10 @@
                 }
             }
         },
-        "revision": "v2.2",
+        "revision": "v2.1",
         "use_imagenet_stats": true,
         "video_backend": "torchcodec",
-        "tolerance_s": 0.02,
+        "tolerance_s": 0.0001,
         "timestamps_check": "warn"
     },
     "env": null,
@@ -120,7 +120,7 @@
         "tokenizer_max_length": 48,
         "num_steps": 10,
         "use_cache": true,
-        "freeze_vision_encoder": true,
+        "freeze_vision_encoder": false,
         "train_expert_only": false,
         "train_state_proj": true,
         "optimizer_lr": 2e-05,
@@ -147,17 +147,17 @@
         "min_period": 0.004,
         "max_period": 4.0
     },
-    "output_dir": "outputs/train/2025-08-15/01-34-33_smolvla",
+    "output_dir": "outputs/train/2025-08-15/03-56-10_smolvla",
     "job_name": "smolvla",
     "resume": false,
     "seed": 1000,
     "num_workers": 4,
-    "batch_size": 32,
-    "steps": 2000,
+    "batch_size": 4,
+    "steps": 40000,
     "eval_freq": 20000,
     "log_freq": 100,
     "save_checkpoint": true,
-    "save_freq": 1000,
+    "save_freq": 40000,
     "use_policy_training_preset": true,
     "optimizer": {
         "type": "adamw",