robotics-diffusion-transformer
commited on
Commit
•
97cdc39
1
Parent(s):
7e4f2e7
Update README.md
Browse files
README.md
CHANGED
@@ -34,21 +34,21 @@ Here's an example of how to use the RDT-1B model for inference on a Mobile-ALOHA
|
|
34 |
```python
|
35 |
# Clone the repository and install dependencies
|
36 |
from scripts.agilex_model import create_model
|
37 |
-
|
|
|
38 |
config = {
|
39 |
'episode_len': 1000, # Max length of one episode
|
40 |
'state_dim': 14, # Dimension of the robot's state
|
41 |
'chunk_size': 64, # Number of actions to predict in one step
|
42 |
'camera_names': CAMERA_NAMES,
|
43 |
}
|
44 |
-
control_frequency=25
|
45 |
pretrained_vision_encoder_name_or_path = "google/siglip-so400m-patch14-384"
|
46 |
# Create the model with specified configuration
|
47 |
model = create_model(
|
48 |
args=config,
|
49 |
dtype=torch.bfloat16,
|
50 |
pretrained_vision_encoder_name_or_path=pretrained_vision_encoder_name_or_path,
|
51 |
-
control_frequency=
|
52 |
)
|
53 |
# Start inference process
|
54 |
# Load pre-computed language embeddings
|
|
|
34 |
```python
|
35 |
# Clone the repository and install dependencies
|
36 |
from scripts.agilex_model import create_model
|
37 |
+
# Names of cameras used for visual input
|
38 |
+
CAMERA_NAMES = ['cam_high', 'cam_right_wrist', 'cam_left_wrist']
|
39 |
config = {
|
40 |
'episode_len': 1000, # Max length of one episode
|
41 |
'state_dim': 14, # Dimension of the robot's state
|
42 |
'chunk_size': 64, # Number of actions to predict in one step
|
43 |
'camera_names': CAMERA_NAMES,
|
44 |
}
|
|
|
45 |
pretrained_vision_encoder_name_or_path = "google/siglip-so400m-patch14-384"
|
46 |
# Create the model with specified configuration
|
47 |
model = create_model(
|
48 |
args=config,
|
49 |
dtype=torch.bfloat16,
|
50 |
pretrained_vision_encoder_name_or_path=pretrained_vision_encoder_name_or_path,
|
51 |
+
control_frequency=25,
|
52 |
)
|
53 |
# Start inference process
|
54 |
# Load pre-computed language embeddings
|