Create low precision model

* - required field

Model

Basic parameters model
model domain NLP {{ domain.name | modelList | titlecase }}
input * {{ input }} {{ inputs.length > 1 && order['input'].indexOf(input) !== -1 && input !== 'custom' ? '(' + (order['input'].indexOf(input) + 1) + ')' : null }} input * No inputs found for this model. output * {{ output }} {{ outputs.length > 1 && order['output'].indexOf(output) !== -1 && output !== 'custom' ? '(' + (order['output'].indexOf(output) + 1) + ')' : null }} output * output * No outputs found for this model.
Order of the inputs and outputs matters.

Dataset

Calibration

provided dataset {{ dataLoader.name }} {{ param.name }} dataset location * Fill the code template before tuning
Default calibration sample size: 100

Evaluation

Use the same data as calibration provided dataset {{ dataLoader.name }} {{ param.name }} dataset location * Fill the code template before tuning

Transforms

transform {{ index }} {{ transformation.name }} {{ param.name }}

Metric

metric {{ metric.name }} Fill the code template before tuning {{ param.name }} {{ option }} {{ param.name }} yes no {{ param.name }}

* - required field

Tune

tuning strategy {{ tuning.name }} accuracy goal objective {{ objective.name }}
timeout max trials random seed

Quantization

approach {{ approach.name }}

Calibration

sampling size

Benchmarking

batch size warmup iteration
cores per instance number of instances cores_per_instance * num_of_instance should not be greater than cores_per_socket ({{ secondFormGroup.get('num_of_instance').value }} * {{ secondFormGroup.get('cores_per_instance').value }} > {{ modelService.systemInfo['cores_per_socket'] }})