Accuracy#
Evaluation Dataset#
Properties: 50,000 validation images from ImageNet dataset
Methodology and KPI#
The performance of zero shot accuracy of NV-CLIP NIM on ImageNet validation dataset.
Model |
top-1 Accuracy |
---|---|
ViT-H-224 |
0.7786 |
The performance of zero shot accuracy of NV-CLIP NIM on ImageNet validation dataset of optimized TensorRT models are as follow:
GPU |
Precision |
top-1 Accuracy |
---|---|---|
H100 SXM |
FP16 |
76.91 |
H100 NVL |
FP16 |
76.92 |
H100 PCIe |
FP16 |
76.92 |
A100 SXM |
FP16 |
76.94 |
A100 PCIe |
FP16 |
76.94 |
L40S |
FP16 |
76.9 |
L4 |
FP16 |
76.93 |
A10G |
FP16 |
76.88 |
A6000 Ada |
FP16 |
76.92 |
RTX 4090 |
FP16 |
76.94 |
RTX 5080-WSL |
FP16 |
76.91 |
RTX 5090-WSL |
FP16 |
76.91 |
GH-200 |
FP16 |
76.92 |