Fix formatting and expand training hyperparameters

README.md

@@ -8,34 +8,33 @@ tags:
 - smolvla
 - robotics
 - ur7e
-- ur7e
 - code-as-policies
 - imitation-learning
 - CoRL2026
 ---
 
-    # SmolVLA UR7e Arrange Block 100epi (10 epochs)
+# SmolVLA UR7e Arrange Block 100epi (10 epochs)
 
-    This repository contains a SmolVLA policy checkpoint fine-tuned with LeRobot. The model card is intentionally detailed so the training run can be reproduced or debugged from the uploaded artifact.
+This repository contains a SmolVLA policy checkpoint fine-tuned with LeRobot. The model card is intentionally detailed so the training run can be reproduced or debugged from the uploaded artifact.
 
-    ## Model Details
+## Model Details
 
-    - **Policy:** SmolVLA
-    - **Base checkpoint:** [`lerobot/smolvla_base`](https://huggingface.co/lerobot/smolvla_base)
-    - **Training dataset:** [`CoRL2026-CSI/UR7e-CaP_arrange_block_100epi`](https://huggingface.co/datasets/CoRL2026-CSI/UR7e-CaP_arrange_block_100epi)
-    - **Training script:** `lerobot/scripts/train_smolvla_ur7e.sh`
-    - **Checkpoint:** step `5520`, approximately `10.00` epochs
-    - **Reported training loss at checkpoint:** `0.009`
-    - **Resolved config:** [`train_config.json`](train_config.json)
+- **Policy:** SmolVLA
+- **Base checkpoint:** [`lerobot/smolvla_base`](https://huggingface.co/lerobot/smolvla_base)
+- **Training dataset:** [`CoRL2026-CSI/UR7e-CaP_arrange_block_100epi`](https://huggingface.co/datasets/CoRL2026-CSI/UR7e-CaP_arrange_block_100epi)
+- **Training script:** `lerobot/scripts/train_smolvla_ur7e.sh`
+- **Checkpoint:** step `5520`, approximately `10.00` epochs
+- **Reported training loss at checkpoint:** `0.009`
+- **Resolved config:** [`train_config.json`](train_config.json)
 
-    Related checkpoints from the same run:
+Related checkpoints from the same run:
 
-    - [5ep checkpoint](https://huggingface.co/CoRL2026-CSI/smolvla_ur7e_arrange_block_100epi_5ep)
+- [5ep checkpoint](https://huggingface.co/CoRL2026-CSI/smolvla_ur7e_arrange_block_100epi_5ep)
 - [10ep checkpoint](https://huggingface.co/CoRL2026-CSI/smolvla_ur7e_arrange_block_100epi_10ep)
 
-    ## Dataset
+## Dataset
 
-    | Key | Value |
+| Key | Value |
 |---|---|
 | `Robot` | UR7e |
 | `Episodes` | 100 |
@@ -45,11 +44,11 @@ tags:
 | `Camera streams` | `observation.images.realsense_wrist`, `observation.images.realsense_topview` |
 | `Dataset state/action shape` | [7] / [7] |
 
-    ## Reproduction
+## Reproduction
 
-    The uploaded [`train_config.json`](train_config.json) is the authoritative serialized LeRobot config for this checkpoint. The table below mirrors the key values for quick inspection.
+The uploaded [`train_config.json`](train_config.json) is the authoritative serialized LeRobot config for this checkpoint. The table below mirrors the key values for quick inspection.
 
-    | Key | Value |
+| Key | Value |
 |---|---|
 | `script` | lerobot/scripts/train_smolvla_ur7e.sh |
 | `job_name` | smolvla_ur7e_arrange_block_100epi_bs64_acc4_ep10_20260509_130552 |
@@ -63,18 +62,18 @@ tags:
 | `checkpoint_lr` | 2.5e-06 |
 | `effective_batch` | 64 x 1 x 4 = 256 |
 
-    Approximate script invocation:
+Approximate script invocation:
 
-    ```bash
-    cd /home/work/hscho/corl_2026/AutoDataCollector/lerobot
+```bash
+cd /home/work/hscho/corl_2026/AutoDataCollector/lerobot
 CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DATASET_REPO_ID="CoRL2026-CSI/UR7e-CaP_arrange_block_100epi" BATCH_SIZE="64" GRADIENT_ACCUMULATION_STEPS="4" STEPS="5520" NUM_WORKERS="4" DATALOADER_PREFETCH_FACTOR="1" CUDA_VISIBLE_DEVICES="0" NUM_GPUS="1" MIXED_PRECISION="bf16" SAVE_FREQ="2760" LOG_FREQ="10" EVAL_FREQ="0" WANDB_PROJECT="lerobot-smolvla-ur7e" OMP_NUM_THREADS="4" MKL_NUM_THREADS="4" PYTORCH_CUDA_ALLOC_CONF="expandable_segments:True" bash train_smolvla_ur7e.sh
-    ```
+```
 
-    ## Detailed Hyperparameters
+## Detailed Hyperparameters
 
-    ### Script Defaults and Environment
+### Script Defaults and Environment
 
-    | Key | Value |
+| Key | Value |
 |---|---|
 | `CONDA_ENV` | lerobot |
 | `POLICY_TYPE` | smolvla |
@@ -96,9 +95,9 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 | `MKL_NUM_THREADS` | 4 |
 | `PYTORCH_CUDA_ALLOC_CONF` | expandable_segments:True |
 
-    ### Training Loop and Dataloader
+### Training Loop and Dataloader
 
-    | Key | Value |
+| Key | Value |
 |---|---|
 | `steps` | 5520 |
 | `batch_size` | 64 |
@@ -115,9 +114,9 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 | `ddp_find_unused_parameters` | True |
 | `profile_timing` | False |
 
-    ### Dataset Pipeline
+### Dataset Pipeline
 
-    | Key | Value |
+| Key | Value |
 |---|---|
 | `dataset.repo_id` | CoRL2026-CSI/UR7e-CaP_arrange_block_100epi |
 | `dataset.root` | `null` |
@@ -127,9 +126,9 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 | `dataset.video_backend` | torchcodec |
 | `dataset.streaming` | False |
 
-    Image augmentation settings:
+Image augmentation settings:
 
-    ```json
+```json
 {
   "enable": true,
   "max_num_transforms": 2,
@@ -203,18 +202,18 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 }
 ```
 
-    Camera rename map:
+Camera rename map:
 
-    ```json
+```json
 {
   "observation.images.realsense_wrist": "observation.images.camera1",
   "observation.images.realsense_topview": "observation.images.camera2"
 }
 ```
 
-    ### Policy Configuration
+### Policy Configuration
 
-    ```json
+```json
 {
   "type": "smolvla",
   "pretrained_path": "lerobot/smolvla_base",
@@ -294,9 +293,9 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 }
 ```
 
-    ### Optimizer
+### Optimizer
 
-    ```json
+```json
 {
   "type": "adamw",
   "lr": 0.0001,
@@ -310,9 +309,9 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 }
 ```
 
-    ### Scheduler
+### Scheduler
 
-    ```json
+```json
 {
   "type": "cosine_decay_with_warmup",
   "num_warmup_steps": 1000,
@@ -322,9 +321,9 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 }
 ```
 
-    ### Logging
+### Logging
 
-    ```json
+```json
 {
   "enable": true,
   "disable_artifact": false,
@@ -336,25 +335,25 @@ CONDA_ENV="lerobot" POLICY_TYPE="smolvla" POLICY_PATH="lerobot/smolvla_base" DAT
 }
 ```
 
-    ## Usage
+## Usage
 
-    Use this model as a LeRobot policy checkpoint:
+Use this model as a LeRobot policy checkpoint:
 
-    ```bash
-    python -m lerobot.scripts.lerobot_eval \
-      --policy.path=CoRL2026-CSI/smolvla_ur7e_arrange_block_100epi_10ep
-    ```
+```bash
+python -m lerobot.scripts.lerobot_eval \
+  --policy.path=CoRL2026-CSI/smolvla_ur7e_arrange_block_100epi_10ep
+```
 
-    For Python loading inside LeRobot code, use the SmolVLA policy loader with this repository id as the pretrained path.
+For Python loading inside LeRobot code, use the SmolVLA policy loader with this repository id as the pretrained path.
 
-    ## Evaluation and Limitations
+## Evaluation and Limitations
 
-    This model card reports training checkpoint information only. No rollout success rate or task-level evaluation metric is included in this repository.
+This model card reports training checkpoint information only. No rollout success rate or task-level evaluation metric is included in this repository.
 
-    The checkpoint assumes a compatible observation/action schema and the camera remapping shown above. The optimizer/RNG `training_state` files are not included; only the loadable `pretrained_model` artifact is uploaded.
+The checkpoint assumes a compatible observation/action schema and the camera remapping shown above. The optimizer/RNG `training_state` files are not included; only the loadable `pretrained_model` artifact is uploaded.
 
-    ## Provenance
+## Provenance
 
-    - VLM backbone: [`HuggingFaceTB/SmolVLM2-500M-Video-Instruct`](https://huggingface.co/HuggingFaceTB/SmolVLM2-500M-Video-Instruct)
-    - Fine-tuning run: `smolvla_ur7e_arrange_block_100epi_bs64_acc4_ep10_20260509_130552`
-    - Source training script: `lerobot/scripts/train_smolvla_ur7e.sh`
+- VLM backbone: [`HuggingFaceTB/SmolVLM2-500M-Video-Instruct`](https://huggingface.co/HuggingFaceTB/SmolVLM2-500M-Video-Instruct)
+- Fine-tuning run: `smolvla_ur7e_arrange_block_100epi_bs64_acc4_ep10_20260509_130552`
+- Source training script: `lerobot/scripts/train_smolvla_ur7e.sh`