Model params 104 MB
Estimates for a single full pass of model at input size 512 x 512:
- Memory required for features: 337 MB
- Flops: 91 GFLOPs
Estimates are given below of the burden of computing the relu4_3 features in the network for different input sizes using a batch size of 128:
| input size | feature size | feature memory | flops |
|---|---|---|---|
| 256 x 256 | 32 x 32 x 512 | 9 GB | 2 TFLOPs |
| 512 x 512 | 64 x 64 x 512 | 36 GB | 9 TFLOPs |
| 768 x 768 | 96 x 96 x 512 | 82 GB | 21 TFLOPs |
| 1024 x 1024 | 128 x 128 x 512 | 146 GB | 37 TFLOPs |
| 1280 x 1280 | 160 x 160 x 512 | 228 GB | 59 TFLOPs |
| 1536 x 1536 | 192 x 192 x 512 | 328 GB | 84 TFLOPs |
A rough outline of where in the network memory is allocated to parameters and features and where the greatest computational cost lies is shown below. The x-axis does not show labels (it becomes hard to read for networks containing hundreds of layers) - it should be interpreted as depicting increasing depth from left to right. The goal is simply to give some idea of the overall profile of the model:
