flops

FLOPs Metrics API

FLOPS calculation for keras.Model or keras.Sequential.

Functions

Name	Description
get_flops	Calculate FLOPS for keras.Model or keras.Sequential

Machine-readable model

• get_flopsfunctionCalculate FLOPS for keras.Model or keras.Sequential.

function

get_flops

Python

Calculate FLOPS for keras.Model or keras.Sequential.

helia_edge/metrics/flops.py:17

get_flops(model: keras.Model, batch_size: int | None = None, fpath: os.PathLike | None = None) -> float

Calculate FLOPS for keras.Model or keras.Sequential. Ignore operations used in only training mode such as Initialization.

Note

Only tensorflow backend is supported currently. Does not support LSTM and GRU.

Example:

model = keras.Sequential([
    keras.layers.Dense(128, activation="relu", input_shape=(784,)),
    keras.layers.Dense(64, activation="relu"),
    keras.layers.Dense(10, activation="softmax"),
])
import helia_edge as helia

model.build()
flops = helia.metrics.get_flops(model, batch_size=1)
print(f"FLOPS: {flops/1e6:.2f}M")

Parameters

Parameters of get_flops

Name	Type	Default	Description
model	keras.Model | keras.Sequential	Required	Model
batch_size	int	None	Batch size. Defaults to None.
fpath	os.PathLike	None	Output file path. Defaults to None.

Returns

Returns of get_flops

Value	Type	Description
float	float	FLOPS