Upload policy weights, train config and readme

README.md

@@ -6,8 +6,8 @@ license: apache-2.0
 model_name: smolvla
 pipeline_tag: robotics
 tags:
-- robotics
 - smolvla
+- robotics
 - lerobot
 ---
 

config.json

@@ -1,6 +1,6 @@
 {
     "type": "smolvla",
-    "n_obs_steps": 1,
+    "n_obs_steps": 2,
     "normalization_mapping": {
         "VISUAL": "IDENTITY",
         "STATE": "MEAN_STD",
@@ -75,7 +75,7 @@
     "scheduler_decay_lr": 2.5e-06,
     "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
     "load_vlm_weights": true,
-    "add_image_special_tokens": false,
+    "add_image_special_tokens": true,
     "attention_mode": "cross_attn",
     "prefix_length": 0,
     "pad_language_to": "max_length",

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f7a028858fda986532adb9ab47240d59f540dcb5e014c2c7b0df7b735dc81ff
+oid sha256:2d21ec3e915b40cef0eaadca1a3bc4a1a6b90b25b7f4d0101e9685c066bf6416
 size 906713296

train_config.json

@@ -67,7 +67,7 @@
     "env": null,
     "policy": {
         "type": "smolvla",
-        "n_obs_steps": 1,
+        "n_obs_steps": 2,
         "normalization_mapping": {
             "VISUAL": "IDENTITY",
             "STATE": "MEAN_STD",
@@ -142,7 +142,7 @@
         "scheduler_decay_lr": 2.5e-06,
         "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
         "load_vlm_weights": true,
-        "add_image_special_tokens": false,
+        "add_image_special_tokens": true,
         "attention_mode": "cross_attn",
         "prefix_length": 0,
         "pad_language_to": "max_length",