Construct a request for a Blade service

Prerequisites

Before you begin, ensure that you have:

• A Blade service deployed on PAI-EAS

• The endpoint URL and service name for your deployment

• Python 3 installed

Install the SDK

pip install -U eas-prediction --user

Call a Blade service

The following example shows the complete request flow: initialize a client, build a BladeRequest with input tensors, send the request, and read the output.

#!/usr/bin/env python
from eas_prediction import PredictClient
from eas_prediction import BladeRequest

# Replace with your actual endpoint URL and service name
client = PredictClient('<endpoint-url>', '<service-name>')
client.init()

# Build the request — add each input tensor with its shape and data type
req = BladeRequest()
req.add_feed('input_data', 1, [1, 360, 128], BladeRequest.DT_FLOAT, [0.8] * 46080)
req.add_feed('input_length', 1, [1], BladeRequest.DT_INT32, [187])
req.add_feed('start_token', 1, [1], BladeRequest.DT_INT32, [104])

# Declare the output tensor to fetch
req.add_fetch('output', BladeRequest.DT_FLOAT)

# Send the request and read the response
resp = client.predict(req)
print(resp.get_tensor_shape('output'))

Replace the following placeholders with your actual values:

Note: If your PAI-EAS service requires token-based authentication, pass the token when initializing the client. For details, see the eas-python-sdk repository.

What's next

• For the full SDK API reference and additional examples, see the eas-python-sdk repository.