GoogLeNet Models

Overview

GoogLeNet introduces a groundbreaking approach to efficiently utilizing computational resources, enabling the network to achieve greater depth and width without exceeding a fixed computational budget. This efficiency is the result of a thoughtfully designed architecture that incorporates multi-scale processing and draws inspiration from Hebbian theory. Comprising 22 layers, GoogLeNet is optimized to deliver high-quality performance in both classification and detection tasks.

Getting Started

Follow these steps to use and convert GoogLeNet models using PyTorch and TorchVision.

1. Install Required Libraries:

   Ensure you have the necessary libraries installed:

   pip install torch torchvision
   

2. Load and Convert GoogLeNet Model:

   Load a pretrained GoogLeNet model using PyTorch and TorchVision, create a dummy input tensor for tracing, trace the model to convert it to TorchScript, and finally save the traced model.

   import torch
   import torchvision
   
   model = torchvision.models.googlenet(pretrained=True)
   trace_data = torch.randn(1, 3, 224, 224)
   trace_model = torch.jit.trace(model.cpu().eval(), trace_data)
   torch.jit.save(trace_model, 'googlenet_float.pt')
   

How It Works ?

Before you begin, ensure that the NeuroPilot Converter Tool is installed.

Quant8 Conversion Process

1. Generate Calibration Data:

   The following script creates a directory named data and generates 100 batches of random input data, each saved as a .npy file. This data is used for calibration during the quantization process.

   import os
   import numpy as np
   
   os.mkdir('data')
   for i in range(100):
       data = np.random.randn(1, 3, 224, 224).astype(np.float32)
       np.save('data/batch_{}.npy'.format(i), data)
   

2. Convert to Quantized TFLite Format:

   Use the following command to convert the model to a quantized TFLite format using the generated calibration data:

   mtk_pytorch_converter                                 \
       --input_script_module_file=googlenet_float.pt    \
       --output_file=googlenet_ptq_quant.tflite         \
       --input_shapes=1,3,224,224                       \
       --quantize=True                                  \
       --input_value_ranges=-1,1                        \
       --calibration_data_dir=data/                     \
       --calibration_data_regexp=batch_.*\.npy
   

FP32 Conversion Process

To convert the model to a non-quantized (FP32) TFLite format, use the following command:

mtk_pytorch_converter                                 \
    --input_script_module_file=googlenet_float.pt     \
    --output_file=googlenet_float.tflite              \
    --input_shapes=1,3,224,224

Model Details

General Information

Property	Value
Category	Classification
Input Size	224x224
GFLOPS	1.50
#Params (M)	6.62
Training Framework	PyTorch
Inference Framework	TFLite
Quant8 Model package	Download
Float32 Model package	Download

Model Properties

Quant8

• Format: TensorFlow Lite v3

• Description: Exported by NeuroPilot converter v7.14.1+release

Inputs

Property	Value
Name	x.2
Tensor	int8[1,3,224,224]
Identifier	151
Quantization	Linear
Quantization Range	-1.0039 ≤ 0.0078 * q ≤ 0.9961

Outputs

Property	Value
Name	1162
Tensor	int8[1,1000]
Identifier	79
Quantization	Linear
Quantization Range	-2.7253 ≤ 0.0373 * (q + 55) ≤ 6.7947

Fp32

• Format: TensorFlow Lite v3

• Description: Exported by NeuroPilot converter v2.9.0

Inputs

Property	Value
Name	x.2
Tensor	float32[1,3,224,224]
Identifier	157

Outputs

Property	Value
Name	1162
Tensor	float32[1,1000]
Identifier	176

Performance Benchmarks

GoogLeNet-quant8

Run model (.tflite) 10 times	CPU (Thread:8)	GPU	ARMNN(GpuAcc)	ARMNN(CpuAcc)	Neuron Stable Delegate(APU)	APU(MDLA)	APU(VPU)
G350	N/A	N/A	N/A	N/A	N/A	N/A	N/A
G510	N/A	N/A	N/A	N/A	N/A	3.04 ms	N/A
G700	N/A	N/A	N/A	N/A	N/A	2.04 ms	N/A
G1200	N/A	N/A	N/A	N/A	N/A	2.39 ms	N/A

GoogLeNet-fp32

Run model (.tflite) 10 times	CPU (Thread:8)	GPU	ARMNN(GpuAcc)	ARMNN(CpuAcc)	Neuron Stable Delegate(APU)	APU(MDLA)	APU(VPU)
G350	193.555 ms (Thread:4)	157.701 ms	133.454 ms	111.099 ms	N/A	N/A	673.224 ms
G510	177.939 ms	52.755 ms	45.865 ms	39.105 ms	8.522 ms	8.4 ms	N/A
G700	35.967 ms	39.538 ms	34.542 ms	33.717 ms	6.224 ms	6.04 ms	N/A
G1200	31.778 ms	28.887 ms	24.254 ms	20.328 ms	7.072 ms	6.05 ms	N/A

• Widespread: CPU only, light workload.

• Performance: CPU and GPU, medium workload.

• Ultimate: CPU, GPU, and APUs, heavy workload.

Resources

github