VGG16 convergence comparison on GPU between fluid and tensorflow

[kuke]

• Data set: cifar10
• Epochs: 50
• Batch size: 128
• Optimizer: Adam
• Learning rate = 1e-3
• Conclusion: Fluid converges a bit faster and they get very close accuracy after adequate training