haiku_trainer Documentation¶
haiku_trainer is a simple python class for training dm-haiku machine learning models.
from functools import partial
import haiku as hk
import jax
import jax.numpy as jnp
import optax
from haiku_trainer import Trainer
# fake data iterator
def data_iter(mode='train'):
rng = jax.random.PRNGKey(42 if mode == 'train' else 43)
x = jax.random.normal(rng, (32, 5))
y = jnp.sum(x, axis=1, keepdims=True) * 9. - 1.
while True:
yield (x, y)
# MLP haiku model
def mlp(x, is_training):
x = hk.Linear(64)(x)
x = jax.nn.relu(x)
x = hk.dropout(hk.next_rng_key(), 0.5, x) if is_training else x
return hk.Linear(1)(x)
# mse loss function
def loss_fn(inputs, is_training):
x, y = inputs
y_hat = mlp(x, is_training=is_training)
return jnp.mean(jnp.square(y_hat - y))
# create trainer object
trainer = Trainer(
train_loss_fn=partial(loss_fn, is_training=True),
train_data_iter=data_iter('train'),
val_loss_fn=partial(loss_fn, is_training=False),
val_data_iter=data_iter('val'),
optimizer=optax.adam(1e-3),
ckpt_freq=5000,
logging_freq=1000,
out_dir='/tmp/regression',
resume=False)
# train model
trainer.fit(total_steps=10_000)
Installation¶
To install the latest version of haiku_trainer, run:
$ pip install git+https://github.com/ntt123/haiku_trainer
API Documentation
License¶
haiku_trainer is licensed under the MIT License.