neural_data_type
How one element of a tensor is stored.
| Name | Meaning |
|---|---|
B_NEURAL_NO_TYPE | 0: no type |
B_NEURAL_FLOAT32 | 32-bit float |
B_NEURAL_FLOAT16 | 16-bit float |
B_NEURAL_INT8 | signed 8-bit integer, usually quantized |
B_NEURAL_UINT8 | unsigned 8-bit integer, usually quantized |
B_NEURAL_INT16 | signed 16-bit integer |
B_NEURAL_INT32 | signed 32-bit integer (shapes, indices) |
B_NEURAL_INT64 | signed 64-bit integer |
B_NEURAL_BOOL | boolean |
