A library to interact with the stdp.io REST API
Project description
stdp.io client library
A library to interact with the stdp.io REST API and a framework to keep your local Akida models synced with your cloud stored models. You can train your models in the cloud and have them synced to your local devices.
authenticating
Authenticate via username and password or via token. Specify the directory you want fetched models to be stored and synced.
from stdpio import stdpio
stdp = stdpio(username="username", password="password", model_dir="/tmp/")
or, if you already have a token
stdp = stdpio(token="example-token", model_dir="/tmp/")
to get a token:
stdp = stdpio(username="username", password="password")
token = stdp.get_token()
fetching your Akida models
This will query the stdp.io API to bring back a list of your models.
#get models
models = stdp.my_models()
# fetch all the model files & knowns
for model in models:
# download the model file from stdp.io
stdp.fetch_model_file(model)
# print the labels and trained neurons for the model
print(stdp.fetch_known(model))
You can also add search paramaters:
params = {"name__icontains": "mobilenet"}
stdp.my_models(**params)
syncing your Akida models
When stdpio is initialised, it will begin a timer that will check models updated_at date. If you want to keep models synced with stdp.io, simply at it to the list of models to sync. The interval defaults to 1 second checks, this can be changed by passing in the keyword argument 'initerval'
stdp = stdpio(token="example-token", model_dir="/tmp/", interval=1)
stdp.sync_model("dbe69029-6ad0-4609-a06b-b0958e892f15")
and to stop syncing a model
stdp.unsync_model("dbe69029-6ad0-4609-a06b-b0958e892f15")
running inference while syncing models
# sync an Akida model
model_key = "058cdf7b-b913-44a9-9fd9-52d41d4295d0"
stdp = stdpio(token="example-token", model_dir="")
stdp.sync_model(model_key)
# loading / fetching the models via API
model_path = False
while not model_path:
model_path = stdp.get_model_path(model_key)
model_object = stdp.get_model_object(model_key)
akida_model = Model(filename=model_path)
while True:
# check for model changes
minted_model_path = stdp.get_model_path(model_key)
if minted_model_path:
akida_model = Model(filename=minted_model_path)
if akida_model:
image = load_img("thor.png")
input_arr = tf.keras.preprocessing.image.img_to_array(image)
input_arr = tf.image.resize_with_crop_or_pad(input_arr, 224, 224)
input_arr = np.array([input_arr], dtype="uint8")
predictions = akida_model.predict(input_arr, num_classes=model["model_output_features"])
time.sleep(1)
to get more information about the model object
model_object = stdp.get_model_object(model_key)
print(model_object)
an example model object will look like this:
{
"unique_id": "058cdf7b-b913-44a9-9fd9-52d41d4295d0",
"name": "testing",
"short_description": "test",
"labels": {
"0": "test",
"3": "mouse"
},
"learned": [
0,
3
],
"base_model": "mobilenet_imagenet",
"updated_at": "2021-10-11T11:27:22.811873Z",
"model_input_shape": "[224, 224, 3]",
"model_output_shape": "[1, 1, 10]",
"model_output_features": 10
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.