Statements (23)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
gptkb:floating_point_unit
|
| gptkbp:advantage |
lower precision than IEEE 754 half-precision
wider exponent range than IEEE 754 half-precision |
| gptkbp:compatibleWith |
gptkb:IEEE_754
|
| gptkbp:distinctFrom |
IEEE 754 half-precision
|
| gptkbp:exponentBits |
8
|
| gptkbp:fullName |
Brain Floating Point 16
|
| gptkbp:introduced |
gptkb:Google
|
| gptkbp:introducedIn |
2018
|
| gptkbp:signBit |
1
|
| gptkbp:significandBits |
7
|
| gptkbp:usedBy |
gptkb:Intel_CPUs
gptkb:NVIDIA_Ampere_GPUs gptkb:AMD_CPUs gptkb:Google_TPU |
| gptkbp:usedFor |
neural network training
neural network inference |
| gptkbp:usedIn |
gptkb:machine_learning
deep learning |
| gptkbp:width |
16
|
| gptkbp:bfsParent |
gptkb:Tensor_Cores
|
| gptkbp:bfsLayer |
7
|
| https://www.w3.org/2000/01/rdf-schema#label |
BFLOAT16
|