Skip to main content
onyx.save_model(
    name: str,
    model: Union[MLP, RNN, Transformer],
    source_datasets: List[Dict[str, Optional[str]]] = []
)
Uploads an Onyx model to the Engine. Typically used for locally-trained models; the Engine automatically saves models it trains.

​
Parameters

​
name
str
required
The name for the model. Must be a non-empty string.
​
model
Union[MLP, RNN, Transformer]
required
The Onyx model to save. Must be an MLP, RNN, or Transformer instance.
​
source_datasets
List[Dict[str, Optional[str]]]
default:"[]"
Source datasets used to train this model, for lineage tracking. Each dictionary should have:
  • name (str): Name of the source dataset
  • version_id (str, optional): Specific version, or latest if not provided

​
Returns

None. Prints a confirmation message when upload completes.

​
Raises

  • Exception: If the name is an empty string
  • Exception: If a source dataset is not found in the Engine

​
Example

from onyxengine import Onyx
from onyxengine.modeling import MLP, MLPConfig, Input, Output

# Initialize the client
onyx = Onyx()

# Create model configuration
outputs = [Output(name='acceleration')]
inputs = [
    Input(name='velocity', parent='acceleration', relation='derivative'),
    Input(name='position', parent='velocity', relation='derivative'),
    Input(name='control_input'),
]

config = MLPConfig(
    outputs=outputs,
    inputs=inputs,
    dt=0.01,
    sequence_length=8,
    hidden_layers=3,
    hidden_size=64
)

# Create and save model
model = MLP(config)
onyx.save_model(
    name='my_model',
    model=model,
    source_datasets=[{'name': 'my_training_data'}]
)

​
Notes

  • The model is saved locally before uploading (in ~/.onyx/models/)
  • Model weights are saved in PyTorch format (.pt)
  • The model configuration is saved alongside the weights
  • Each save creates a new version of the model