The following examples use .set width 175, unless overridden in the query:
SELECT * FROM mldb.ONNXSeq2Seq(
ON (SELECT id, content as txt, headline, category FROM JulesBelvezeDummyData WHERE id <= 5) as InputTable
ON (SELECT * FROM onnx_models WHERE model_id='t5-small-headline-generator') AS ModelTable DIMENSION
ON (SELECT tokenizer FROM sequence_tokenizers WHERE model_id='t5-small-headline-generator') AS TokenizerTable DIMENSION
USING
Accumulate('id')
ModelOutputTensor('sequences')
Const_min_length(10)
Const_max_length(84)
Const_num_beams(4)
Const_repetition_penalty(1.2)
Const_length_penalty(2.0)
Const_num_return_sequences(1)
) AS td
ORDER BY id;
*** Query completed. 5 rows found. 2 columns returned.
*** Total elapsed time was 5 seconds.
id sequences ------ ----------------------------------------------------------------------------- 1 Facebook releases a report that shows what content was most widely viewed by Americans between April and June 2 'Quantum Battery': The Bigger the Battery, The Faster It Charged, demonstrating Superabsorption at Work 3 Founder Elon Musk laughed off the incident, saying it was fine 4 Scalene is a high-performance CPU and memory profiler for Python 5 'Cube Satellites' are taking photos and infrared images of factories, retail parking lots, oil tankers, and more and selling them to hedge funds and banks looking