Inference Data Schema Definitions

Concepts regarding definitions for submitting inference data to Wallaroo.

Table of Contents

Inference requests sent to Wallaroo are submitted under the inference data schema definitions. The following details those requirements for:

  • Submitting single inference requests versus batch inference requests
  • Submitting inference requests with multidimensional array data.

Inferencing Data Types

Wallaroo pipelines accept either Apache Arrow tables natively or pandas DataFrames.

  • If using the Wallaroo SDK with wallaroo.pipeline.infer or wallaroo.pipeline.infer_from_file, it is a “What you give is what you get” situation: If an Apache Arrow table is submitted for an inference request with the , then an Apache Arrow table is returned. If a pandas DataFrame is submitted, then a pandas DataFrame is returned.
  • If using the Wallaroo pipeline’s inference URL as an API request, the API headers Content-Type and Accept determine what data type is submitted and what data type is received:
    • For DataFrame formatted JSON: application/json; format=pandas-records
    • For Arrow binary files: application/vnd.apache.arrow.file

Data Types

Data types for inputs and outputs to inference requests through Wallaroo is based on the Apache Arrow Data Types, with the following exceptions:

  • null: Not allowed. All fields must have submitted values that match their data type. For example, if the schema expects a float value, then some value of type float such as 0.0 must be submitted and cannot be None or Null. If a schema expects a string value, then some value of type string must be submitted, etc. The exception are BYOP models, which can accept optional inputs.
  • time32 and time64: Datetime data types must be converted to string.

Data Constraints

Data submitted to Wallaroo for inference requests have the following data constraints.

  • Equal rows constraint: The number of input rows and output rows must match.
  • Data Type Consistency: Data types within each tensor are of the same type.

Equal Rows

Each input row for an inference is related directly to the inference row output.

For example, the INPUT and OUTPUT rows match, with each input row directly corresponding to an output row.

Equal Rows Input Example

 tensor
0[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
1[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
2[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
3[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
4[0.5817662108, 0.09788155100000001, 0.1546819424, 0.4754101949, -0.19788623060000002, -0.45043448540000003, 0.016654044700000002, -0.0256070551, 0.0920561602, -0.2783917153, 0.059329944100000004, -0.0196585416, -0.4225083157, -0.12175388770000001, 1.5473094894000001, 0.2391622864, 0.3553974881, -0.7685165301, -0.7000849355000001, -0.1190043285, -0.3450517133, -1.1065114108, 0.2523411195, 0.0209441826, 0.2199267436, 0.2540689265, -0.0450225094, 0.10867738980000001, 0.2547179311]

Equal Rows Output Example

 timein.tensorout.dense_1anomaly.count
02023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
12023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
22023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
32023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
42023-11-17 20:34:17.005[0.5817662108, 0.097881551, 0.1546819424, 0.4754101949, -0.1978862306, -0.4504344854, 0.0166540447, -0.0256070551, 0.0920561602, -0.2783917153, 0.0593299441, -0.0196585416, -0.4225083157, -0.1217538877, 1.5473094894, 0.2391622864, 0.3553974881, -0.7685165301, -0.7000849355, -0.1190043285, -0.3450517133, -1.1065114108, 0.2523411195, 0.0209441826, 0.2199267436, 0.2540689265, -0.0450225094, 0.1086773898, 0.2547179311][0.0010916889]0

Data Type Consistency

Each element must have the same internal data type. For example, the following is valid because all of the data types within each element are float32.

t = [
    [2.35, 5.75],
    [3.72, 8.55],
    [5.55, 97.2]
]

The following is invalid, as it mixes floats and strings in each element:

t = [
    [2.35, "Bob"],
    [3.72, "Nancy"],
    [5.55, "Wani"]
]

The following inputs are valid, as each data type is consistent within the elements.

df = pd.DataFrame({
    "t": [
        [2.35, 5.75, 19.2],
        [5.55, 7.2, 15.7],
    ],
    "s": [
        ["Bob", "Nancy", "Wani"],
        ["Jason", "Rita", "Phoebe"]
    ]
})
df
 ts
0[2.35, 5.75, 19.2][Bob, Nancy, Wani]
1[5.55, 7.2, 15.7][Jason, Rita, Phoebe]

Data Types

Data types for inputs and outputs to inference requests through Wallaroo is based on the Apache Arrow Data Types, with the following exceptions:

  • null: Not allowed. All fields must have submitted values that match their data type. For example, if the schema expects a float value, then some value of type float such as 0.0 must be submitted and cannot be None or Null. If a schema expects a string value, then some value of type string must be submitted, etc. The exception are BYOP models, which can accept optional inputs.
  • time32 and time64: Datetime data types must be converted to string.

Data Constraints

Data submitted to Wallaroo for inference requests have the following data constraints.

  • Equal rows constraint: The number of input rows and output rows must match.
  • Data Type Consistency: Data types within each tensor are of the same type.

Equal Rows

Each input row for an inference is related directly to the inference row output.

For example, the INPUT and OUTPUT rows match, with each input row directly corresponding to an output row.

Equal Rows Input Example

 tensor
0[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
1[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
2[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
3[-1.0603297501, 2.3544967095000002, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192000001, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526000001, 1.9870535692, 0.7005485718000001, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439]
4[0.5817662108, 0.09788155100000001, 0.1546819424, 0.4754101949, -0.19788623060000002, -0.45043448540000003, 0.016654044700000002, -0.0256070551, 0.0920561602, -0.2783917153, 0.059329944100000004, -0.0196585416, -0.4225083157, -0.12175388770000001, 1.5473094894000001, 0.2391622864, 0.3553974881, -0.7685165301, -0.7000849355000001, -0.1190043285, -0.3450517133, -1.1065114108, 0.2523411195, 0.0209441826, 0.2199267436, 0.2540689265, -0.0450225094, 0.10867738980000001, 0.2547179311]

Equal Rows Output Example

 timein.tensorout.dense_1anomaly.count
02023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
12023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
22023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
32023-11-17 20:34:17.005[-1.0603297501, 2.3544967095, -3.5638788326, 5.1387348926, -1.2308457019, -0.7687824608, -3.5881228109, 1.8880837663, -3.2789674274, -3.9563254554, 4.0993439118, -5.6539176395, -0.8775733373, -9.131571192, -0.6093537873, -3.7480276773, -5.0309125017, -0.8748149526, 1.9870535692, 0.7005485718, 0.9204422758, -0.1041491809, 0.3229564351, -0.7418141657, 0.0384120159, 1.0993439146, 1.2603409756, -0.1466244739, -1.4463212439][0.99300325]0
42023-11-17 20:34:17.005[0.5817662108, 0.097881551, 0.1546819424, 0.4754101949, -0.1978862306, -0.4504344854, 0.0166540447, -0.0256070551, 0.0920561602, -0.2783917153, 0.0593299441, -0.0196585416, -0.4225083157, -0.1217538877, 1.5473094894, 0.2391622864, 0.3553974881, -0.7685165301, -0.7000849355, -0.1190043285, -0.3450517133, -1.1065114108, 0.2523411195, 0.0209441826, 0.2199267436, 0.2540689265, -0.0450225094, 0.1086773898, 0.2547179311][0.0010916889]0

Data Type Consistency

Each element must have the same internal data type. For example, the following is valid because all of the data types within each element are float32.

t = [
    [2.35, 5.75],
    [3.72, 8.55],
    [5.55, 97.2]
]

The following is invalid, as it mixes floats and strings in each element:

t = [
    [2.35, "Bob"],
    [3.72, "Nancy"],
    [5.55, "Wani"]
]

The following inputs are valid, as each data type is consistent within the elements.

df = pd.DataFrame({
    "t": [
        [2.35, 5.75, 19.2],
        [5.55, 7.2, 15.7],
    ],
    "s": [
        ["Bob", "Nancy", "Wani"],
        ["Jason", "Rita", "Phoebe"]
    ]
})
df
 ts
0[2.35, 5.75, 19.2][Bob, Nancy, Wani]
1[5.55, 7.2, 15.7][Jason, Rita, Phoebe]

Inference Requests Examples

The following examples show different inference requests with data sets.

Inference Request Batch Inputs

For standard batch inputs, each column and row corresponds to one data input. For example, a set of house data fed into a house price model has the following inputs.

 number_of_bedroomsyear_builtnumber_of_bathrooms
031990-06-011
142023-07-022.5
232000-08-151.5

As the data types meet all three data requirements, the model is set with the configuration batch_config=None, allowing for batch inference submissions. The following represents a inference results from the data above for a house price model, with each output row corresponding to the input row.

 timein.number_of_bedroomsin.year_builtin.number_of_bathroomsout.variable
02023-10-31 16:57:28.97431990-06-011[137079.4]
12023-10-31 16:57:28.97442023-07-022.5[2002393.6]
22023-10-31 16:57:28.97432000-08-151.5[1967344.1]

Inference Request Batch Inputs for a Range of Data

For batch inputs, each inference request is contained in a single row of data. For example, a bike rental model predicts rentals based on previous dates for bikes sold, whether it was a week day, etc. These are submitted as a group of dates that are then used by the model to predict bike rentals on figure dates. For example, this 5 day period is submitted for the inference request.

 dtedaysite_idcntseasonholidayweekdayworkingday
02011-02-02site000112401031
12011-02-03site000115511041
22011-02-04site000123241051
32011-02-05site00018051060
42011-02-06site000119481000

To submit to the Wallaroo deployed pipeline as a batch, each range of dates and values is set into a single row of data. This is reflected in the input and output schemas defined for the model. The pyarrow schema pa.list_ is used to define the data being received as a List per field.

import pyarrow as pa
import wallaroo

wl = wallaroo.Client()

input_schema = pa.schema([
   pa.field('dteday', pa.list_(pa.string()) ),
   pa.field('site_id', pa.list_(pa.string()) ), 
   pa.field('cnt', pa.list_(pa.int64()) ),
   pa.field('season', pa.list_(pa.int64()) ),
   pa.field('holiday', pa.list_(pa.int64()) ),
   pa.field('weekday', pa.list_(pa.int64()) ),
   pa.field('workingday', pa.list_(pa.int64()) ),           
])

output_schema = pa.schema([
    pa.field('dteday', pa.list_(pa.string()) ),
    pa.field('site_id', pa.list_(pa.string()) ), 
    pa.field('forecast', pa.list_(pa.int64()) ),
])

arima_model = (wl.upload_model('bikeforecast-arima', 
                               './models/forecast.py', 
                               framework=Framework.PYTHON)
                               .configure(runtime="python", 
                               input_schema = input_schema,
                               output_schema = output_schema
                               )
              )

Here, a month of data for single site is prepared for inference through a Wallaroo deployed model. Each inference request contains in a single row per site. Additional sites would be in their own row, with their month’s worth of data set as lists within the cells.

display(input_frame)
 dtedaysite_idcntseasonholidayweekdayworkingday
0[2011-02-02, 2011-02-03, 2011-02-04, 2011-02-05, 2011-02-06, 2011-02-07, 2011-02-08, 2011-02-09, 2011-02-10, 2011-02-11, 2011-02-12, 2011-02-13, 2011-02-14, 2011-02-15, 2011-02-16, 2011-02-17, 2011-02-18, 2011-02-19, 2011-02-20, 2011-02-21, 2011-02-22, 2011-02-23, 2011-02-24, 2011-02-25, 2011-02-26, 2011-02-27, 2011-02-28, 2011-03-01, 2011-03-02, 2011-03-03, 2011-03-04, 2011-03-05, 2011-03-06, 2011-03-07, 2011-03-08][site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001, site0001][1240, 1551, 2324, 805, 1948, 1650, 913, 931, 1256, 1614, 1000, 1883, 1964, 2036, 2586, 3219, 3947, 1826, 1418, 723, 1281, 2564, 2181, 1539, 2059, 2428, 836, 1235, -1, -1, -1, -1, -1, -1, -1][1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1][0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0][3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2][1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1]

The inference results are then converted to whatever format the data scientist and ML engineers require.

 timeout.dtedayout.forecastout.site_idanomaly.count
02023-09-22 00:15:20.214[2011-03-02, 2011-03-03, 2011-03-04, 2011-03-05, 2011-03-06, 2011-03-07, 2011-03-08][2269, 1712, 1795, 1371, 1819, 2045, 1974][site0001, site0001, site0001, site0001, site0001, site0001, site0001]0

Using Multi-Dimensional Numpy Arrays for Inference Inputs

Wallaroo inference requests support either pandas DataFrame or Apache Arrows as inputs. When using a numpy array with multi-dimensional shapes, the Wallaroo SDK provides helper functions for data conversion to support the Wallaroo data requirements.

  • Wallaroo SDK:
    • DataFrame: Numpy multi-dimensional array native support.
    • Apache Arrow: Numpy multi-dimensional array must be flattened.
  • Wallaroo MLOps API:
    • DataFrame: Numpy multi-dimensional array must be flattened.
    • Apache Arrow: Numpy multi-dimensional array must be flattened.

Note that when using the Wallaroo MLOps with multi-dimensional numpy arrays, the helper functions are contained in the Wallaroo SDK to assist with data conversion.

Using Multi-Dimensional Numpy Arrays with the Wallaroo SDK

Inference requests made through the Wallaroo SDK accept Numpy arrays as input values within a pandas DataFrame. For Apache Arrow tables, the numpy array must be flattened.

The Wallaroo SDK method wallaroo.utils.flatten_np_array_columns(df, col) converts a pandas DataFrame column specified with the name col. The pandas DataFrame is then converted to an Apache Arrow table for inferencing.

Wallaroo SDK Numpy Arrays with DataFrames Example

The following example deploys a Computer Vision Mobilenet model. A multi-dimensional numpy array is generated from an image, then used it directly for inference request in a DataFrame.

import wallaroo
import torch
import cv2

wl = wallaroo.Client()

# upload the model
mobilenet_model = wl.upload_model("mobilenet-demo", 
                './models/mobilenet.pt.onnx', 
                framework=Framework.ONNX).configure(batch_config="single")

# build the pipeline with the model and deploy
pipeline = wl.build_pipeline("cv-sample-pipeline")
pipeline.add_model_step(mobilenet_model)
# deployed the pipeline
pipeline.deploy()

# convert the image to a tensor
image = cv2.imread("./data/images/input/example/dairy_bottles.png")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
tensor = torch.FloatTensor(image)

# get the numpy value and use in a dataframe
npArray = tensor.cpu().numpy()
df_with_numpy = pd.DataFrame({"tensor":[npArray]})

The following is an example of the the first row and the first 4 columns from the image converted into a multi-dimensional numpy array.

 tensors
0[[0.9372549, 0.9529412, 0.9490196], [0.93333334, 0.9490196, 0.9490196], [0.9254902, 0.9529412, 0.94509804], [0.91764706, 0.9490196, 0.9490196]]

The DataFrame with the multi-dimensional numpy array is submitted to the Wallaroo pipeline in an inference request. The following isolates the first 4 objects detected and overall confidence value of all detected objects.

infResults = pipeline.infer(df_with_numpy, timeout=300)
df = pd.DataFrame([
    {"time": infResults.loc[0, ['time']][0],
    'out.avg_conf': [infResults.loc[0, ['out.avg_conf']][0]],
    'boxes': [infResults.loc[0, ['out.boxes']][0][0:4][0:4]],
    'classes': [infResults.loc[0, ['out.classes']][0][0:4]],
    'confidences': [infResults.loc[0, ['out.confidences']][0][0:4]]
    }
])
display(df)
 timeout.avg_confboxesclassesconfidences
02024-02-07 15:22:21.497[0.2895055][[[0.0, 210.2901, 85.26464, 479.07495], [72.037796, 197.3227, 151.44221, 468.43225], [211.28014, 184.72838, 277.21924, 420.42746], [143.23906, 203.83008, 216.85551, 448.8881]]][[44, 44, 44, 44]][[0.98649, 0.9011532, 0.60778517, 0.59223205]]

Wallaroo SDK Numpy Arrays with Apache Arrow Example

Apache Arrow tables submitted for inference request must have multi-dimensional arrays flattened before they are submitted. The following example shows flattening the numpy array with the wallaroo.utils.flatten_np_array_columns(df, col) method, which returns the column’s values as flattened arrays.

import wallaroo
import torch
import cv2

wl = wallaroo.Client()

# upload the model
mobilenet_model = wl.upload_model("mobilenet-demo", 
                './models/mobilenet.pt.onnx', 
                framework=Framework.ONNX).configure(batch_config="single")

# build the pipeline with the model and deploy
pipeline = wl.build_pipeline("cv-sample-pipeline")
pipeline.add_model_step(mobilenet_model)
# deployed the pipeline
pipeline.deploy()

# convert the image to a tensor
image = cv2.imread("./data/images/input/example/dairy_bottles.png")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
tensor = torch.FloatTensor(image)

# get the numpy value and use in a dataframe
npArray = tensor.cpu().numpy()
df_with_numpy = pd.DataFrame({"tensor":[npArray]})

The following shows the the first 3 columns of the first 4 rows of the image file converted to tensors as a multi-dimensional numpy array in a pandas DataFrame.

 tensors
0[[0.9372549, 0.9529412, 0.9490196], [0.93333334, 0.9490196, 0.9490196], [0.9254902, 0.9529412, 0.94509804], [0.91764706, 0.9490196, 0.9490196]]

Before submitting as an Apache Arrow table, the multi-dimensional arrays are flattened.

The following demonstrates using the wallaroo.utils.flatten_np_array_columns(df, col) takes a pandas DataFrame, and the column name of the arrays to flatten method. For this example, we will flatten the column and retrieve a new array from it.

wallaroo.utils.flatten_np_array_columns(df_with_numpy, 'tensor')

0    [0.9372549, 0.9529412, 0.9490196, 0.94509804, 0.94509804, 0.94509804 ...]
Name: tensor, dtype: object

We create an Apache Arrow table from the flattened numpy array and submit it for the inference request.

table = pa.table([
    pa.array(wallaroo.utils.flatten_np_array_columns(df_with_numpy, 'tensor'))
], names=["tensor"])

display(table)

pyarrow.Table
tensor: list<item: float>
  child 0, item: float
----
tensor: [[[0.9372549,0.9529412,0.9490196,0.94509804,...]]]

result = pipeline.infer(table)

display(result['out.avg_confidence'])

<pyarrow.lib.ChunkedArray object at 0x2a69b4b30>
[
  [
    [
      0.3588039
    ]
  ]
]

Using Multi-Dimensional Numpy Arrays with the Wallaroo MLOps API

The Wallaroo MLOps API supports the following Content-Type and Accept headers values:

  • application/json; format=pandas-records: JSON in pandas Record format.
  • application/vnd.apache.arrow.file: A binary in Apache Arrow format.

In both cases, multi-dimensional numpy array must be flattened before submitting it as part of an API Inference Request.

The following examples demonstrates using the Wallaroo SDK wallaroo.utils.flatten_np_array_columns(df, col) method to flatten a column in a pandas DataFrame that contains multi-dimensional numpy arrays.

Wallaroo MLOps API Numpy Arrays with DataFrames Example

The following is an example of a DataFrame with an image converted to a multi-dimensional numpy array of tensors.

 tensors
0[[0.9372549, 0.9529412, 0.9490196], [0.93333334, 0.9490196, 0.9490196], [0.9254902, 0.9529412, 0.94509804], [0.91764706, 0.9490196, 0.9490196]]

Before submitting as an MLOps API request as a pandas DataFrame, the multi-dimensional arrays are flattened. The following demonstrates using the wallaroo.utils.flatten_np_array_columns(df, col) method. The flattened array is inserted as the tensor column for a new pandas DataFrame.

df_with_flattened_array = pd.DataFrame({'tensor': wallaroo.utils.flatten_np_array_columns(df_with_numpy, 'tensor')})

# show only the first 5 tensors
df = df_with_flattened_array.apply(lambda x: [i[0:5] for i in x])
display(df)
 tensor
0[0.9372549, 0.9529412, 0.9490196, 0.94509804, 0.94509804]

With the multi-dimensional arrays flattened, it is submitted as an inference request.

headers = wl.auth.auth_header()

headers['Content-Type'] = 'application/json; format=pandas-records'

deploy_url = pipeline._deployment._url()

response = requests.post(
                    deploy_url, 
                    headers=headers, 
                    data=df_with_flattened_array.to_json(orient="records")
                )

display(pd.DataFrame(response.json()).loc[:, ['time', 'out.avg_confidence']])
 timeout
01705687869292{‘avg_confidence’: [0.3588039]}

Wallaroo MLOps API Numpy Arrays with Apache Arrow Example

The following is an example of a DataFrame with an image converted to a multi-dimensional numpy array of tensors.

 tensors
0[[0.9372549, 0.9529412, 0.9490196], [0.93333334, 0.9490196, 0.9490196], [0.9254902, 0.9529412, 0.94509804], [0.91764706, 0.9490196, 0.9490196]]

Before submitting as an MLOps API request as an Apache Arrow table, the multi-dimensional arrays are flattened. The following demonstrates using the wallaroo.utils.flatten_np_array_columns(df, col) method.

wallaroo.utils.flatten_np_array_columns(df_with_numpy, 'tensor')

0    [0.9372549, 0.9529412, 0.9490196, 0.94509804, 0.94509804, 0.94509804 ...]
Name: tensor, dtype: object

With the multi-dimensional arrays flattened, it is inserted into an Apache Arrow table and submitted in an inference request.

table = pa.table([
    pa.array(wallaroo.utils.flatten_np_array_columns(df_with_numpy, 'tensor'))
], names=["tensor"])

display(table)

pyarrow.Table
tensor: list<item: float>
  child 0, item: float
----
tensor: [[[0.9372549,0.9529412,0.9490196,0.94509804,...]]]

headers = wl.auth.auth_header()

# set Content-Type type
headers['Content-Type']='application/vnd.apache.arrow.file'

# set accept as apache arrow table
headers['Accept']="application/vnd.apache.arrow.file"

deploy_url = pipeline._deployment._url()

response = requests.post(
                    deploy_url, 
                    headers=headers, 
                    data=table, 
                    verify=True
                )

# Arrow table is retrieved 
with pa.ipc.open_file(response.content) as reader:
    arrow_table = reader.read_all()

display(result['out.avg_confidence'])

<pyarrow.lib.ChunkedArray object at 0x2a69b4b30>
[
  [
    [
      0.3588039
    ]
  ]
]

Tutorials

Other examples come from the Wallaroo Tutorials as listed below.