Candidate File Generation

Build a scored candidate file from your logs, then evaluate it.

1 Choose Mode

Upload Precomputed Scores

You already have a file with item scores from your model.

Score via External API Coming Soon

Call an external scoring API for each user-item pair.

Score via ONNX Model Not Available

Requires pip install onnxruntime

Historical Logs (required)

CSV or Parquet. Must contain user_id and item_id columns.

Decision ID Column — auto-detected if present

Identifies each event when the same (user, item) appears multiple times.

Candidate Scores File (required)

Must contain user_id, item_id, and score columns.

ONNX Model Configuration

ONNX Model File *

.onnx format, max 500 MB.

Input Feature Mapping

Map ONNX input names to columns in your logs.

→

Output Name

Score Index

Batch Size

Max Rows

Feature Source