concepts.dsl.tensor_value.simple_quantize_tvalue#
- simple_quantize_tvalue(value)[source]#
Quantize a value into a quantized value using the simple rules.
- The simple rules are:
If the value is already quantized, return the value itself.
If the value has tensor_quantized_values, return the tensor_indices.
- If the value has intrinsically quantized dtypes, quantize the value.
For BOOL, quantize the value with (value > 0.5).to(torch.int64).
For INT64, quantize the value with value.to(torch.int64).
- Parameters:
value (TensorValue) – the value to be quantized.
- Returns:
the quantized value.
- Return type: