first-generation Transformer Engine
GPTKB entity
Statements (17)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
AI hardware component
|
| gptkbp:announced |
2022
|
| gptkbp:developedBy |
gptkb:NVIDIA
|
| gptkbp:enables |
higher throughput for AI workloads
reduced memory usage for AI models |
| gptkbp:feature |
automated precision scaling
dynamic mixed-precision computation |
| https://www.w3.org/2000/01/rdf-schema#label |
first-generation Transformer Engine
|
| gptkbp:introducedIn |
gptkb:NVIDIA_Hopper_architecture
|
| gptkbp:purpose |
accelerate transformer model inference
accelerate transformer model training |
| gptkbp:supports |
FP16 precision
FP8 precision BF16 precision |
| gptkbp:usedIn |
gptkb:NVIDIA_H100_GPU
|
| gptkbp:bfsParent |
gptkb:second-generation_Transformer_Engine
|
| gptkbp:bfsLayer |
7
|