File-Based Model Type

Use the file-based model type to upload, download, or get information on files not supported by one of the format-aware file types in the FrogML SDK.

Upload a File-Based Model to a Machine Learning Repository

You can upload a model to a Machine Learning repository using the frogml.log_model() function. A model can be a single file or a list of files in a directory. This function uses checksum upload, assigning a SHA2 value to each model for retrieval from storage. If the checksum fails, FrogML implements regular upload.Checksum-Based Storage

After uploading the model, FrogML generates a file named model-manifest.json, which contains the model name and its related files and dependencies.

To upload a file-based model to a Machine Learning repository, use the following function:

import frogml
frogml.files.log_model(   repository="<REPOSITORY_KEY>",    # The JFrog repository to upload the model to.
   namespace="<NAMESPACE_NAME>",     # Optional. The namespace or organization.
   model_name="<MODEL_NAME>",     # The uploaded model name
   version="<MODEL_VERSION>",     # Optional. The uploaded model version
   source_path="<FILE_PATH>",     # The file path of the model on your machine.
   properties={<PROPERTIES>}    # Optional. These are properties you can set on the model to categorize and label it.
   dependencies=[<DEPENDENCY>, <DEPENDENCY>],  # Optional
   code_path= <CODE_PATH>
)

Parameters

Example

import frogml
frogml.files.log_model(
repository="frog-ml-local",
namespace="frog-ml-beta",     # Optional
model_name="my-cool-model",
version="1.0.0",    # Optional
source_path="~/root/models/my-cool-model/",
properties={"model_type": "keras", "experiment": "my-exp"} # Optional
dependencies=["numpy>=1.19"],  # Optional
code_path= "./src" #Optional
)

Download a File-Based Model from a Machine Learning Repository

You can use the frogml.load_model() function to download a file-based model from an Artifactory repository. This method will download the model file locally and return the path to the returned model.

import frogml

# Download model version
frogml.files.load_model(
   repository="<REPOSITORY_KEY>",  # The name of the JFrog repository you want to download the model from
   namespace="<NAMESPACE_NAME>",   # Optional. The name of the namespace or organization: use this parameter to group models by project or group.
   model_name="<MODEL_NAME>",      # The name of the model you want to download.
   version="<MODEL_VERSION>",      # The version of the model you want to download.
   target_path="<FILE_PATH>",      # The file path to where you want to download the model.
)

Parameters

Example

import frogml

# Download model version
frogml.files.load_model(
    repository="frog-ml-local",
    namespace="frog-ml-beta",     #optional
    model_name="my-cool-model",
    version="1.0.0"
    target_path="~/root/models/",
)

Return Value

The load_model() method returns the path to the local downloaded model as a Path object imported from pathlib.

Get Information on a File-Based Model in a Machine Learning Repository

You can use the frogml.get_model_info() function to retrieve information on a specific file-based model version without downloading the full model files.

import frogml

# Download model version
frogml.files.get_model_info(
   repository="<REPOSITORY_KEY>",  # The name of the JFrog repository you want to download the model from
   namespace="<NAMESPACE_NAME>",   # Optional. The name of the namespace or organization: use this parameter to group models by project or group.
   model_name="<MODEL_NAME>",      # The name of the model you want to download.
   version="<MODEL_VERSION>",      # The version of the model you want to download.
  
)

Parameters

Example

import frogml

# Download model version
frogml.files.get_model_info(
    repository="frog-ml-local",
    namespace="frog-ml-beta",     #optional
    model_name="my-cool-model",
    version="1.0.0",
    
)

Return Values

{'framework': 'files',
'framework_version': ",
'runtime': 'python',
'runtime_version': '3.9.6', 
'serialization_format': 'pkl'}

[{“artifact_path”: “<artifactory_path>”,
“checksum”: “<file_checksum>”,
“download_path”:”<download_url>”}]

[{“artifact_path”: “<artifactory_path>”,
“checksum”: “<file_checksum>”,
“download_path”:”<download_url>”}]

[{“artifact_path”: “<artifactory_path>”,
“checksum”: “<file_checksum>”,
“download_path”:”<download_url>”}]

e.g. 2024-11-12T13:25:53.20