alexandersoare
commited on
Commit
•
f7872bf
1
Parent(s):
e6d30a8
Upload folder using huggingface_hub
Browse files
README.md
CHANGED
@@ -9,6 +9,8 @@ pipeline_tag: robotics
|
|
9 |
Diffusion Policy (as per [Diffusion Policy: Visuomotor Policy
|
10 |
Learning via Action Diffusion](https://arxiv.org/abs/2303.04137)) trained for the `PushT` environment from [gym-pusht](https://github.com/huggingface/gym-pusht) with keypoint-only observations.
|
11 |
|
|
|
|
|
12 |
## How to Get Started with the Model
|
13 |
|
14 |
Use `python lerobot/scripts/eval.py -p lerobot/diffusion_pusht` to evaluate for 50 episodes with the outputs sent to `outputs/eval`.
|
@@ -17,9 +19,9 @@ For further information, please see the [LeRobot library](https://github.com/hug
|
|
17 |
|
18 |
## Training Details
|
19 |
|
20 |
-
Trained with TODO commit hash
|
21 |
|
22 |
-
The model was trained using [LeRobot's training script](
|
23 |
|
24 |
```bash
|
25 |
python lerobot/scripts/train.py \
|
@@ -55,7 +57,7 @@ The model was evaluated on the `PushT` environment from [gym-pusht](https://gith
|
|
55 |
Here are the metrics for 500 episodes worth of evaluation.
|
56 |
|
57 |
Metric|Average over 500 episodes
|
58 |
-
|
59 |
Average max. overlap ratio | 0.97
|
60 |
Success rate (%) | 71.0
|
61 |
|
|
|
9 |
Diffusion Policy (as per [Diffusion Policy: Visuomotor Policy
|
10 |
Learning via Action Diffusion](https://arxiv.org/abs/2303.04137)) trained for the `PushT` environment from [gym-pusht](https://github.com/huggingface/gym-pusht) with keypoint-only observations.
|
11 |
|
12 |
+
Note: The original work trains keypoints-only with conditioning via inpainting. Here, we encode the observation along with the agent position and use the encoding as global conditioning for the denoising U-Net.
|
13 |
+
|
14 |
## How to Get Started with the Model
|
15 |
|
16 |
Use `python lerobot/scripts/eval.py -p lerobot/diffusion_pusht` to evaluate for 50 episodes with the outputs sent to `outputs/eval`.
|
|
|
19 |
|
20 |
## Training Details
|
21 |
|
22 |
+
Trained with TODO commit hash one the PR is merged
|
23 |
|
24 |
+
The model was trained using [LeRobot's training script](TODO link with commit hash one the PR is merged) and with the [pusht_keypoints](https://huggingface.co/datasets/lerobot/pusht_keypoints/tree/v1.5) dataset, using this command:
|
25 |
|
26 |
```bash
|
27 |
python lerobot/scripts/train.py \
|
|
|
57 |
Here are the metrics for 500 episodes worth of evaluation.
|
58 |
|
59 |
Metric|Average over 500 episodes
|
60 |
+
-|-
|
61 |
Average max. overlap ratio | 0.97
|
62 |
Success rate (%) | 71.0
|
63 |
|