How to count the allocated memory size of the model

TensorFlow General Discussion
,
1 
  When we use the benchmark test, it will display   "peak memeroy footprint (MB) :init =   overall =   "     but the size value displayed in each test is different.
 We would like to know is there any other way to count the memory size of the model

thanks

3

You can use the model profiler that tells you the model’s memory requirement, no. of parameters, flops etc

Sample code using Toy Example:


from tensorflow.keras.applications import EfficientNetB3

from model_profiler import model_profiler

model = EfficientNetB3(include_top=True)

Batch_size = 128

profile = model_profiler(model, Batch_size)

print(profile)

Output:

Downloading data from https://storage.googleapis.com/keras-applications/efficientnetb3.h5
50102272/50095040 [==============================] - 1s 0us/step
50110464/50095040 [==============================] - 1s 0us/step
| Model Profile                    | Value         | Unit    |
|----------------------------------|---------------|---------|
| Selected GPUs                    | None Detected | GPU IDs |
| No. of FLOPs                     | 1.5651        | BFLOPs  |
| GPU Memory Requirement           | 41.7385       | GB      |
| Model Parameters                 | 12.3205       | Million |
| Memory Required by Model Weights | 46.9991       | MB      |
1 Like
5 
   thanks, Actually we want to count the memory usage of the tflite model when it runs on Android.

Is there any way?