Add README

README.md

@@ -0,0 +1,72 @@
+---
+license: apache-2.0
+base_model: nvidia/GR00T-N1.5-3B
+tags:
+  - robot-learning
+  - gr00t
+  - diffusion-policy
+  - libero
+  - object-centric
+library_name: gr00t
+pipeline_tag: robotics
+inference: false
+---
+
+# Tacoin GR00T Libero Object 8K (Checkpoint 8000)
+
+Tacoin fine-tuned GR00T checkpoint trained on the LIBERO libero object 8k benchmark. The policy consumes dual RGB views (`video.image`, `video.wrist_image`) plus an 8-D state and predicts 16 joint-space actions using the diffusion head.
+
+## Training Snapshot
+
+- **Base model:** `nvidia/GR00T-N1.5-3B`
+- **Checkpoint step:** 8000 / 8000
+- **Dataset:** libero_object (10 tasks, 454 demos @ 10.0 FPS)
+- **Run notes:** 8k-step continuation fine-tune
+
+## Evaluation
+
+Offline reconstruction evaluated on 10 evenly spaced trajectories (160 steps each) with decord backend and `denoising_steps=4`. Metrics are on unnormalized actions.
+
+| Metric | Value |
+| --- | --- |
+| Average MSE | **0.03396** |
+| Median MSE | 0.03179 |
+| Std MSE | 0.01547 |
+| Max MSE | 0.06431 |
+| Fraction ≤ 0.05 | 80.0% |
+| Fraction ≤ 0.075 | 100.0% |
+| Fraction ≤ 0.10 | 100.0% |
+
+## Usage
+
+```python
+from gr00t.experiment.data_config import load_data_config
+from gr00t.model.policy import Gr00tPolicy
+
+ckpt = 'Tacoin/GR00T-N1.5-3B-LIBERO-OBJECT-8K'
+data_config = load_data_config('libero_gr00t')
+policy = Gr00tPolicy(
+    model_path=ckpt,
+    modality_config=data_config.modality_config(),
+    modality_transform=data_config.transform(),
+    embodiment_tag='new_embodiment',
+    denoising_steps=4,
+)
+```
+
+Pass a LeRobot observation dict to `policy.get_action(...)` to obtain the 16-step plan.
+
+## Files
+
+| Path | Description |
+| --- | --- |
+| `config.json` | Transformer config for the action head. |
+| `model-0000x-of-00002.safetensors` | Sharded weights. |
+| `model.safetensors.index.json` | Weight shard index. |
+| `experiment_cfg/metadata.json` | Dataset statistics for normalization. |
+| `optimizer.pt`, `scheduler.pt`, `rng_state.pth` | Optimizer state for resuming. |
+| `trainer_state.json` | Trainer snapshot (loss curves, etc.). |
+
+## License
+
+Apache-2.0. Please credit NVIDIA Isaac GR00T and LIBERO when using this checkpoint.