Client$log()
Batch Logging - Designed for sending batches of data to Arize
The Client$log() is designed for training, validation or production environment where batches of data are processed. These environments may be either a R Studio Notebook or a R server that is batch processing lots of backend data.
Import and initialize Arize R client from the Arize
Client$new()
to call Client$log()
with a R data.frame() containing inference data. ORGANIZATION_KEY <- 'ORGANIZATION_KEY'
API_KEY <- 'API_KEY'
arize_client <- Client$new(
organization_key = ORGANIZATION_KEY,
api_key = API_KEY)
schema <- create_schema(
prediction_id_column_name = "prediction_id",
prediction_label_column_name = "prediction_label",
prediction_score_column_name = "prediction_score",
actual_label_column_name = "actual_label",
actual_score_column_name = "actual_score",
feature_column_names = features,
timestamp_column_name = "prediction_ts"
)
# send training data
arize_client$log(
.data_frame = df_train,
.schema = schema
.model_id = model_id,
.model_version = model_version,
.model_type = model_types$SCORE_CATEGORICAL,
.environment = environments$TRAINING,
)
Parameter | Data Type | Description | Required |
---|---|---|---|
.data_frame | data.frame | data.frame to log | Required |
.schema | arize::create_schema | the schema (see ?arize::create_schema ) | Required |
.model_id | character | character, id for the model | Required |
.model_type | integer | 1 for binary, 2 for numeric, 3 for categorical, 4 for score-categorical | Required |
.environment | environment | 1 for production, 2 for validation, 3 for training | Required |
.model_version | character | character, the model version | Optional |
.batch_id | character | character, the batch id | Optional |
.sync | logical | logical, whether to sync | Optional |
.validate | logical | logical, whether to run validation checks | Optional |
.path | character | character, path to use for serialization | Optional |
Attribute | Data Type | Description | Required |
prediction_id_column_name | character | Column name for prediction_id | Required |
feature_column_names | List[character] | List of column names for features | Optional |
prediction_label_column_name | character | Column name for prediction label | Optional |
prediction_score_column_name | character | Column name for prediction scores | Optional |
actual_label_column_name | character | Column name for actual label | Optional |
actual_score_column_name | str | Column name for numeric sequences. Used for NDCG calculations in ranking models | Optional |
timestamp_column_name | character | Column name for timestamps | Optional |
model_id <- "click_through_rate_categorical_vignette_R" # This is the model name that will show up in Arize
model_version <- "v1.0" # Version of model - can be any string
schema <- create_schema(
prediction_id_column_name = "id",
feature_column_names = features,
prediction_label_column_name = "predictions",
prediction_score_column_name = "CTR_predicted",
actual_label_column_name = "actuals",
actual_score_column_name = "CTR",
timestamp_column_name = "model_date"
)
arize_client$log(
.data_frame = df_train,
.model_id = model_id,
.model_version = model_version,
.model_type = model_types$SCORE_CATEGORICAL,
.environment = environments$TRAINING,
.schema = schema
)
Last modified 1yr ago