for batch_size in batch_sizes:
times = []
N = min(30, int(np.floor((n_events_to_load - 1) / batch_size)))
print("This time, N = ", N)
# Make pedictions and time it
for i in range(N):
print(f"On step {i}/{N}...")
data_tmp = data[(i) * batch_size + 1:(i + 1) * batch_size, :, :, :]
#data_tmp = data_tmp[np.newaxis, :, :, :]
print("Shape of data_tmp:", data_tmp.shape)
data_tmp = np.array(data_tmp, dtype=np.float32)
if i == 0:
interpreter.resize_tensor_input(0, [data_tmp.shape[0], 5, 512, 1])
interpreter.allocate_tensors()
t0 = time.time()
interpreter.set_tensor(input_details[0]['index'], data_tmp)
interpreter.invoke()
t = time.time() - t0
if i != 0:
times.append(t)
print(times)
mean = np.mean(times)
