Upload policy weights, train config and readme

README.md

@@ -1,24 +1,63 @@
-# eval3_sanitycheck
-
-SmolVLA sanity-check overfit policy for Eval 3 coke-can placement.
-
-- Base VLM: `HuggingFaceTB/SmolVLM2-500M-Video-Instruct`
-- Policy type: `smolvla`
-- Dataset: `robot-learning-group47/eval3_overfit20`
-- Local dataset root: `/data/lerobot_datasets/robot-learning-group47/eval3_overfit20`
-- Source episodes: `0,1,6,7,12,13,18,19,24,25,30,31,36,37,42,43,48,49,54,55`
-- Steps: `3000`
-- Batch size: `8`
-- Image augmentation: disabled
-- Trainable params: expert/state/action heads only, vision/VLM frozen
-- Final training loss: approximately `0.182`
-
-Important inference settings are stored in `config.json`:
-
-- `chunk_size=50`
-- `n_action_steps=50`
-- `num_steps=10`
-- `attention_mode=self_attn`
-- `num_vlm_layers=8`
-- `num_expert_layers=4`
-- `resize_imgs_with_padding=[512,512]`
+---
+base_model: lerobot/smolvla_base
+datasets: robot-learning-group47/eval3_sanity
+library_name: lerobot
+license: apache-2.0
+model_name: smolvla
+pipeline_tag: robotics
+tags:
+- smolvla
+- robotics
+- lerobot
+---
+
+# Model Card for smolvla
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+[SmolVLA](https://huggingface.co/papers/2506.01844) is a compact, efficient vision-language-action model that achieves competitive performance at reduced computational costs and can be deployed on consumer-grade hardware.
+
+
+This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
+See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
+
+---
+
+## How to Get Started with the Model
+
+For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
+Below is the short version on how to train and run inference/eval:
+
+### Train from scratch
+
+```bash
+lerobot-train \
+  --dataset.repo_id=${HF_USER}/<dataset> \
+  --policy.type=act \
+  --output_dir=outputs/train/<desired_policy_repo_id> \
+  --job_name=lerobot_training \
+  --policy.device=cuda \
+  --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
+  --wandb.enable=true
+```
+
+_Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
+
+### Evaluate the policy/run inference
+
+```bash
+lerobot-record \
+  --robot.type=so100_follower \
+  --dataset.repo_id=<hf_user>/eval_<dataset> \
+  --policy.path=<hf_user>/<desired_policy_repo_id> \
+  --episodes=10
+```
+
+Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
+
+---
+
+## Model Details
+
+- **License:** apache-2.0

config.json

@@ -12,8 +12,8 @@
             "type": "VISUAL",
             "shape": [
                 3,
-                240,
-                320
+                256,
+                256
             ]
         }
     },
@@ -28,12 +28,12 @@
     "device": "cuda",
     "use_amp": false,
     "use_peft": false,
-    "push_to_hub": false,
-    "repo_id": null,
+    "push_to_hub": true,
+    "repo_id": "robot-learning-group47/eval3_sanity",
     "private": null,
     "tags": null,
     "license": null,
-    "pretrained_path": null,
+    "pretrained_path": "lerobot/smolvla_base",
     "chunk_size": 50,
     "n_action_steps": 50,
     "normalization_mapping": {
@@ -53,8 +53,8 @@
     "tokenizer_max_length": 48,
     "num_steps": 10,
     "use_cache": true,
-    "freeze_vision_encoder": true,
-    "train_expert_only": true,
+    "freeze_vision_encoder": false,
+    "train_expert_only": false,
     "train_state_proj": true,
     "optimizer_lr": 0.0001,
     "optimizer_betas": [
@@ -63,18 +63,18 @@
     ],
     "optimizer_eps": 1e-08,
     "optimizer_weight_decay": 1e-10,
-    "optimizer_grad_clip_norm": 10,
+    "optimizer_grad_clip_norm": 10.0,
     "scheduler_warmup_steps": 100,
     "scheduler_decay_steps": 3000,
     "scheduler_decay_lr": 2.5e-06,
     "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
     "load_vlm_weights": true,
     "add_image_special_tokens": false,
-    "attention_mode": "self_attn",
-    "prefix_length": -1,
-    "pad_language_to": "longest",
-    "num_expert_layers": 4,
-    "num_vlm_layers": 8,
+    "attention_mode": "cross_attn",
+    "prefix_length": 0,
+    "pad_language_to": "max_length",
+    "num_expert_layers": 0,
+    "num_vlm_layers": 16,
     "self_attn_every_n_layers": 2,
     "expert_width_multiplier": 0.75,
     "min_period": 0.004,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6de6bca545855eeae75b82f4b6728fbfdcfbaf85e6e5e3e30fe6ab9f5549d05f
-size 599749280
+oid sha256:0ae144a7212881c51f4d1007ed5735b7108d5cf45ecf0113b04e5ee21cbbdc30
+size 906712520

train_config.json

@@ -1,7 +1,7 @@
 {
     "dataset": {
-        "repo_id": "robot-learning-group47/eval3_overfit20",
-        "root": "/data/lerobot_datasets/robot-learning-group47/eval3_overfit20",
+        "repo_id": "robot-learning-group47/eval3_sanity",
+        "root": null,
         "episodes": null,
         "image_transforms": {
             "enable": false,
@@ -74,9 +74,9 @@
                 }
             }
         },
-        "revision": null,
+        "revision": "main",
         "use_imagenet_stats": true,
-        "video_backend": "torchcodec",
+        "video_backend": "pyav",
         "return_uint8": false,
         "streaming": false
     },
@@ -95,8 +95,8 @@
                 "type": "VISUAL",
                 "shape": [
                     3,
-                    240,
-                    320
+                    256,
+                    256
                 ]
             }
         },
@@ -111,12 +111,12 @@
         "device": "cuda",
         "use_amp": false,
         "use_peft": false,
-        "push_to_hub": false,
-        "repo_id": null,
+        "push_to_hub": true,
+        "repo_id": "robot-learning-group47/eval3_sanity",
         "private": null,
         "tags": null,
         "license": null,
-        "pretrained_path": null,
+        "pretrained_path": "lerobot/smolvla_base",
         "chunk_size": 50,
         "n_action_steps": 50,
         "normalization_mapping": {
@@ -136,8 +136,8 @@
         "tokenizer_max_length": 48,
         "num_steps": 10,
         "use_cache": true,
-        "freeze_vision_encoder": true,
-        "train_expert_only": true,
+        "freeze_vision_encoder": false,
+        "train_expert_only": false,
         "train_state_proj": true,
         "optimizer_lr": 0.0001,
         "optimizer_betas": [
@@ -146,18 +146,18 @@
         ],
         "optimizer_eps": 1e-08,
         "optimizer_weight_decay": 1e-10,
-        "optimizer_grad_clip_norm": 10,
+        "optimizer_grad_clip_norm": 10.0,
         "scheduler_warmup_steps": 100,
         "scheduler_decay_steps": 3000,
         "scheduler_decay_lr": 2.5e-06,
         "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
         "load_vlm_weights": true,
         "add_image_special_tokens": false,
-        "attention_mode": "self_attn",
-        "prefix_length": -1,
-        "pad_language_to": "longest",
-        "num_expert_layers": 4,
-        "num_vlm_layers": 8,
+        "attention_mode": "cross_attn",
+        "prefix_length": 0,
+        "pad_language_to": "max_length",
+        "num_expert_layers": 0,
+        "num_vlm_layers": 16,
         "self_attn_every_n_layers": 2,
         "expert_width_multiplier": 0.75,
         "min_period": 0.004,
@@ -167,27 +167,27 @@
         "compile_mode": "max-autotune"
     },
     "reward_model": null,
-    "output_dir": "outputs/train/2026-05-20/20-23-02_smolvla",
-    "job_name": "smolvla",
+    "output_dir": "outputs/train/eval3_sanity_smolvla_20260522_224000",
+    "job_name": "eval3_sanity_smolvla",
     "resume": false,
     "seed": 1000,
     "cudnn_deterministic": false,
     "num_workers": 4,
-    "batch_size": 8,
+    "batch_size": 32,
     "prefetch_factor": 4,
     "persistent_workers": true,
     "steps": 3000,
-    "eval_freq": 20000,
+    "eval_freq": 0,
     "log_freq": 50,
     "tolerance_s": 0.0001,
     "save_checkpoint": true,
-    "save_freq": 500,
+    "save_freq": 1000,
     "use_policy_training_preset": true,
     "optimizer": {
         "type": "adamw",
         "lr": 0.0001,
         "weight_decay": 1e-10,
-        "grad_clip_norm": 10,
+        "grad_clip_norm": 10.0,
         "betas": [
             0.9,
             0.95
@@ -203,7 +203,7 @@
     },
     "eval": {
         "n_episodes": 50,
-        "batch_size": 11,
+        "batch_size": 19,
         "use_async_envs": true
     },
     "wandb": {