InstantMLDocs

Keras

Log Keras training runs to InstantML with a callback that records epoch metrics from model.fit.

Open .md

Pass im.InstantMLKerasCallback to model.fit(...) and it logs your epoch metrics to InstantML. It is a native Keras callback — no wrapper around your model or training loop.

Install

bash
python -m pip install "instantml[frameworks]"

Authenticate with `instantml login` or INSTANTML_API_KEY if you haven't already.

Add the callback to model.fit

This complete example trains a small dense network on synthetic data:

python
import instantml as im
import keras
import numpy as np

x_train = np.random.rand(512, 16).astype("float32")
y_train = (x_train.sum(axis=1) > 8).astype("float32")

model = keras.Sequential([
    keras.Input(shape=(16,)),
    keras.layers.Dense(32, activation="relu"),
    keras.layers.Dense(1, activation="sigmoid"),
])
model.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])

callback = im.InstantMLKerasCallback(
    project="keras-demo",
    name="dense-baseline",
    config={"optimizer": "adam", "layers": 2},
)

model.fit(
    x_train,
    y_train,
    epochs=10,
    validation_split=0.2,
    callbacks=[callback],
)

Open the project at instantml.ai to see loss, accuracy, and their val_ counterparts charted per epoch.

The callback creates the run on on_train_begin, logs scalar metrics from each on_epoch_end (with step=epoch), and finishes the run on on_train_end. Extra keyword arguments are forwarded to im.init(...); when no project is given it defaults to "keras".

Log per-batch metrics

By default the callback logs once per epoch. Enable per-batch logging with log_batch=True; batch metrics are namespaced under batch/:

python
callback = im.InstantMLKerasCallback(project="keras-demo", log_batch=True)

Reuse an existing run

Pass an already-created run when you want the callback to share a run with other code instead of creating its own:

python
run = im.init(project="keras-demo", name="dense-baseline")
callback = im.InstantMLKerasCallback(run=run)

Next steps