1.8.1

1.8.1

Bug Fixes

• Registry: Fix a bug that caused unsupported model type error while logging a sklearn model with score_samples
  inference method.
• Registry: Fix a bug that model inference service creation fails on an existing and suspended service.

Behavior Change

New Features

• ML Job (PrPr): Update Container Runtime image version to 1.0.1
• ML Job (PrPr): Add enable_metrics argument to job submission APIs to enable publishing service metrics to Event Table.
  See Accessing Event Table service metrics
  for retrieving published metrics
  and Costs of telemetry data collection
  for cost implications.
• Registry: When creating a copy of a ModelVersion with log_model, raise an exception if unsupported arguments are provided.

1.8.0

1.8.0

Bug Fixes

• Modeling: Fix a bug in some metrics that allowed an unsupported version of numpy to be installed
  automatically in the stored procedure, resulting in a numpy error on execution
• Registry: Fix a bug that leads to incorrect Model is does not have _is_inference_api error message when assigning
  a supported model as a property of a CustomModel.
• Registry: Fix a bug that inference is not working when models with more than 500 input features
  are deployed to SPCS.

Behavior Change

• Registry: With FeatureGroupSpec support, auto inferred model signature for transformers.Pipeline models have been
  updated, including:

  • Signature for fill-mask task has been changed from

    ModelSignature(
        inputs=[
            FeatureSpec(name="inputs", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureSpec(name="outputs", dtype=DataType.STRING),
        ],
    )

    to

    ModelSignature(
        inputs=[
            FeatureSpec(name="inputs", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureGroupSpec(
                name="outputs",
                specs=[
                    FeatureSpec(name="sequence", dtype=DataType.STRING),
                    FeatureSpec(name="score", dtype=DataType.DOUBLE),
                    FeatureSpec(name="token", dtype=DataType.INT64),
                    FeatureSpec(name="token_str", dtype=DataType.STRING),
                ],
                shape=(-1,),
            ),
        ],
    )

  • Signature for token-classification task has been changed from

    ModelSignature(
        inputs=[
            FeatureSpec(name="inputs", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureSpec(name="outputs", dtype=DataType.STRING),
        ],
    )

    to

    ModelSignature(
        inputs=[FeatureSpec(name="inputs", dtype=DataType.STRING)],
        outputs=[
            FeatureGroupSpec(
                name="outputs",
                specs=[
                    FeatureSpec(name="word", dtype=DataType.STRING),
                    FeatureSpec(name="score", dtype=DataType.DOUBLE),
                    FeatureSpec(name="entity", dtype=DataType.STRING),
                    FeatureSpec(name="index", dtype=DataType.INT64),
                    FeatureSpec(name="start", dtype=DataType.INT64),
                    FeatureSpec(name="end", dtype=DataType.INT64),
                ],
                shape=(-1,),
            ),
        ],
    )

  • Signature for question-answering task when top_k is larger than 1 has been changed from

    ModelSignature(
        inputs=[
            FeatureSpec(name="question", dtype=DataType.STRING),
            FeatureSpec(name="context", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureSpec(name="outputs", dtype=DataType.STRING),
        ],
    )

    to

    ModelSignature(
        inputs=[
            FeatureSpec(name="question", dtype=DataType.STRING),
            FeatureSpec(name="context", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureGroupSpec(
                name="answers",
                specs=[
                    FeatureSpec(name="score", dtype=DataType.DOUBLE),
                    FeatureSpec(name="start", dtype=DataType.INT64),
                    FeatureSpec(name="end", dtype=DataType.INT64),
                    FeatureSpec(name="answer", dtype=DataType.STRING),
                ],
                shape=(-1,),
            ),
        ],
    )

  • Signature for text-classification task when top_k is None has been changed from

    ModelSignature(
        inputs=[
            FeatureSpec(name="text", dtype=DataType.STRING),
            FeatureSpec(name="text_pair", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureSpec(name="label", dtype=DataType.STRING),
            FeatureSpec(name="score", dtype=DataType.DOUBLE),
        ],
    )

    to

    ModelSignature(
        inputs=[
            FeatureSpec(name="text", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureSpec(name="label", dtype=DataType.STRING),
            FeatureSpec(name="score", dtype=DataType.DOUBLE),
        ],
    )

  • Signature for text-classification task when top_k is not None has been changed from

    ModelSignature(
        inputs=[
            FeatureSpec(name="text", dtype=DataType.STRING),
            FeatureSpec(name="text_pair", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureSpec(name="outputs", dtype=DataType.STRING),
        ],
    )

    to

    ModelSignature(
        inputs=[
            FeatureSpec(name="text", dtype=DataType.STRING),
        ],
        outputs=[
            FeatureGroupSpec(
                name="labels",
                specs=[
                    FeatureSpec(name="label", dtype=DataType.STRING),
                    FeatureSpec(name="score", dtype=DataType.DOUBLE),
                ],
                shape=(-1,),
            ),
        ],
    )

  • Signature for text-generation task has been changed from

    ModelSignature(
        inputs=[FeatureSpec(name="inputs", dtype=DataType.STRING)],
        outputs=[
            FeatureSpec(name="outputs", dtype=DataType.STRING),
        ],
    )

    to

    ModelSignature(
        inputs=[
            FeatureGroupSpec(
                name="inputs",
                specs=[
                    FeatureSpec(name="role", dtype=DataType.STRING),
                    FeatureSpec(name="content", dtype=DataType.STRING),
                ],
                shape=(-1,),
            ),
        ],
        outputs=[
            FeatureGroupSpec(
                name="outputs",
                specs=[
                    FeatureSpec(name="generated_text", dtype=DataType.STRING),
                ],
                shape=(-1,),
            )
        ],
    )

• Registry: PyTorch and TensorFlow models now expect a single tensor input/output by default when logging to Model
  Registry. To use multiple tensors (previous behavior), set options={"multiple_inputs": True}.

  Example with single tensor input:

  import torch
  
  class TorchModel(torch.nn.Module):
      def __init__(self, n_input: int, n_hidden: int, n_out: int, dtype: torch.dtype = torch.float32) -> None:
          super().__init__()
          self.model = torch.nn.Sequential(
              torch.nn.Linear(n_input, n_hidden, dtype=dtype),
              torch.nn.ReLU(),
              torch.nn.Linear(n_hidden, n_out, dtype=dtype),
              torch.nn.Sigmoid(),
          )
  
      def forward(self, tensor: torch.Tensor) -> torch.Tensor:
          return cast(torch.Tensor, self.model(tensor))
  
  # Sample usage:
  data_x = torch.rand(size=(batch_size, n_input))
  
  # Log model with single tensor
  reg.log_model(
      model=model,
      ...,
      sample_input_data=data_x
  )
  
  # Run inference with single tensor
  mv.run(data_x)

  For multiple tensor inputs/outputs, use:

  reg.log_model(
      model=model,
      ...,
      sample_input_data=[data_x_1, data_x_2],
      options={"multiple_inputs": True}
  )

• Registry: Default enable_explainability to False when the model can be deployed to Snowpark Container Services.

New Features

• Registry: Added support to single torch.Tensor, tensorflow.Tensor and tensorflow.Variable as input or output
  data.
• Registry: Support xgboost.DMatrix
  datatype for XGBoost models.

1.7.5

1.7.5

• Support Python 3.12.
• Explainability: Support native and snowml sklearn pipeline

Bug Fixes

• Registry: Fixed a compatibility issue when using snowflake-ml-python 1.7.0 or greater to save a tensorflow.keras
  model with keras 2.x, if relax_version is set or default to True, and newer version of snowflake-ml-python
  is available in Snowflake Anaconda Channel, model could not be run in Snowflake. If you have such model, you could
  use the latest version of snowflake-ml-python and call ModelVersion.load to load it back, and re-log it.
  Alternatively, you can prevent this issue by setting relax_version=False when saving the model.
• Registry: Removed the validation that disallows data that does not have non-null values being passed to
  ModelVersion.run.
• ML Job (PrPr): No longer require CREATE STAGE privilege if stage_name points to an existing stage
• ML Job (PrPr): Fixed a bug causing some payload source and entrypoint path
  combinations to be erroneously rejected with
  ValueError(f"{self.entrypoint} must be a subpath of {self.source}")
• ML Job (PrPr): Fixed a bug in Ray cluster startup config which caused certain Runtime APIs to fail

Behavior Change

New Features

• Registry: Added support for handling Hugging Face model configurations with auto-mapping functionality.
• Registry: Added support for keras 3.x model with tensorflow and pytorch backend
• ML Job (PrPr): Support any serializable (pickleable) argument for @remote decorated functions

1.7.4

1.7.4

• FileSet: The snowflake.ml.fileset.FileSet has been deprecated and will be removed in a future version.
  Use snowflake.ml.dataset.Dataset and
  snowflake.ml.data.DataConnector
  instead.

Bug Fixes

• Registry: Fixed an issue that the hugging face pipeline is loaded using incorrect dtype.
• Registry: Fixed an issue that only 1 row is used when infer the model signature in the modeling model.

Behavior Changes

• Registry: ModelVersion.run on a service would require redeploying the service once account opts into nested function.

New Features

• Add new snowflake.ml.jobs preview API for running headless workloads on SPCS using
  Container Runtime for ML
• Added guardrails option to Cortex complete function, enabling
  Cortex Guard support

1.7.3

1.7.3

• Added lowercase versions of Cortex functions, added deprecation warning to Capitalized versions.
• Bumped the requirements of fsspec and s3fs to >=2024.6.1,<2026
• Bumped the requirement of mlflow to >=2.16.0, <3
• Registry: Support 500+ features for model registry

Bug Fixes

• Registry: Fixed a bug when providing non-range index pandas DataFrame as the input to a ModelVersion.run.
• Registry: Improved random model version name generation to prevent collisions.
• Registry: Fix an issue when inferring signature or running inference with Snowpark data that has a column whose type
  is ARRAY and contains NULL value.
• Registry: ModelVersion.run now accepts fully qualified service name.
• Monitoring: Fix issue in SDK with creating monitors using fully qualified names.
• Registry: Fix error in log_model for any sklearn models with only data pre-processing including pre-processing only
  pipeline models due to default explainability enablement.

Behavior Changes

New Features

• Added user_files argument to Registry.log_model for including images or any extra file with the model.
• Registry: Added support for handling Hugging Face model configurations with auto-mapping functionality
• DataConnector: Add new DataConnector.from_sql() constructor

1.7.2

1.7.2

Bug Fixes

• Model Explainability: Fix issue that explain is enabled for scikit-learn pipeline
  whose task is UNKNOWN and fails later when invoked.

Behavior Changes

New Features

• Registry: Support asynchronous model inference service creation with the block option
  in ModelVersion.create_service() set to True by default.

1.7.1

1.7.1

Bug Fixes

• Registry: Null value is now allowed in the dataframe used in model signature inference. Null values will be ignored
  and others will be used to infer the signature.
• Registry: Pandas Extension DTypes (pandas.StringDType(), pandas.BooleanDType(), etc.) are now supported in model
  signature inference.
• Registry: Null value is now allowed in the dataframe used to predict.
• Data: Fix missing snowflake.ml.data.* module exports in wheel
• Dataset: Fix missing snowflake.ml.dataset.* module exports in wheel.
• Registry: Fix the issue that tf_keras.Model is not recognized as keras model when logging.

Behavior Changes

New Features

• Registry: Option to enable_monitoring set to False by default. This will gate access to preview features of Model Monitoring.
• Model Monitoring: show_model_monitors Registry method. This feature is still in Private Preview.
• Registry: Support pd.Series in input and output data.
• Model Monitoring: add_monitor Registry method. This feature is still in Private Preview.
• Model Monitoring: resume and suspend ModelMonitor. This feature is still in Private Preview.
• Model Monitoring: get_monitor Registry method. This feature is still in Private Preview.
• Model Monitoring: delete_monitor Registry method. This feature is still in Private Preview.

1.7.0

1.7.0

Behavior Change

• Generic: Require python >= 3.9.
• Data Connector: Update to_torch_dataset and to_torch_datapipe to add a dimension for scalar data.
  This allows for more seamless integration with PyTorch DataLoader, which creates batches by stacking inputs of each batch.

Examples:

ds = connector.to_torch_dataset(shuffle=False, batch_size=3)

• Input: "col1": [10, 11, 12]

  • Previous batch: array([10., 11., 12.]) with shape (3,)
  • New batch: array([[10.], [11.], [12.]]) with shape (3, 1)

• Input: "col2": [[0, 100], [1, 110], [2, 200]]

  • Previous batch: array([[ 0, 100], [ 1, 110], [ 2, 200]]) with shape (3,2)
  • New batch: No change

• Model Registry: External access integrations are optional when creating a model inference service in
  Snowflake >= 8.40.0.

• Model Registry: Deprecate build_external_access_integration with build_external_access_integrations in
  ModelVersion.create_service().

Bug Fixes

• Registry: Updated log_model API to accept both signature and sample_input_data parameters.
• Feature Store: ExampleHelper uses fully qualified path for table name. change weather features aggregation from 1d to 1h.
• Data Connector: Return numpy array with appropriate object type instead of list for multi-dimensional
  data from to_torch_dataset and to_torch_datapipe
• Model explainability: Incompatibility between SHAP 0.42.1 and XGB 2.1.1 resolved by using latest SHAP 0.46.0.

New Features

• Registry: Provide pass keyworded variable length of arguments to class ModelContext. Example usage:

mc = custom_model.ModelContext(
    config = 'local_model_dir/config.json',
    m1 = model1
)

class ExamplePipelineModel(custom_model.CustomModel):
    def __init__(self, context: custom_model.ModelContext) -> None:
      super().__init__(context)
      v = open(self.context['config']).read()
      self.bias = json.loads(v)['bias']

    @custom_model.inference_api
    def predict(self, input: pd.DataFrame) -> pd.DataFrame:
      model_output = self.context['m1'].predict(input)
      return pd.DataFrame({'output': model_output + self.bias})

• Model Development: Upgrade scikit-learn in UDTF backend for log_loss metric. As a result, eps argument is now ignored.
• Data Connector: Add the option of passing a None sized batch to to_torch_dataset for better
  interoperability with PyTorch DataLoader.
• Model Registry: Support pandas.CategoricalDtype
• Registry: It is now possible to pass signatures and sample_input_data at the same time to capture background
  data from explainablity and data lineage.

1.6.4

1.6.4

Bug Fixes

• Registry: Fix an issue that leads to incident when using ModelVersion.run with service.

1.6.3

1.6.3

• Model Registry (PrPr) has been removed.

Bug Fixes

• Registry: Fix a bug that when package whose name does not follow PEP-508 is provided when logging the model,
  an unexpected normalization is happening.
• Registry: Fix not a valid remote uri error when logging mlflow models.
• Registry: Fix a bug that ModelVersion.run is called in a nested way.
• Registry: Fix an issue that leads to log_model failure when local package version contains parts other than
  base version.

New Features

• Data: Improve DataConnector.to_pandas() performance when loading from Snowpark DataFrames.
• Model Registry: Allow users to set a model task while using log_model.
• Feature Store: FeatureView supports ON_CREATE or ON_SCHEDULE initialize mode.