1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
|
import ssl
ssl._create_default_https_context = (
ssl._create_unverified_context
) # Fixed problem with downloading CIFAR10 dataset
from flax import linen as nn
import torch
import einx
import os
import jax
import optax
import time
import torchvision
import torchvision.transforms as transforms
import jax.numpy as jnp
from flax.training import train_state
import einx.nn.flax as einn
from functools import partial
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
])
batch_size = 256
rng = jax.random.PRNGKey(42)
def next_rng():
global rng
rng, x = jax.random.split(rng)
return x
cifar10_path = os.path.join(os.path.dirname(__file__), "cifar10")
trainset = torchvision.datasets.CIFAR10(
root=cifar10_path, train=True, download=True, transform=transform
)
trainloader = torch.utils.data.DataLoader(
trainset, batch_size=batch_size, shuffle=True, num_workers=2
)
testset = torchvision.datasets.CIFAR10(
root=cifar10_path, train=False, download=True, transform=transform
)
testloader = torch.utils.data.DataLoader(
testset, batch_size=batch_size, shuffle=False, num_workers=2
)
class Net(nn.Module):
@nn.compact
def __call__(self, x, training):
for c in [1024, 512, 256]:
x = einn.Linear("b [...->c]", c=c)(x)
x = einn.Norm("[b] c", decay_rate=0.99)(x, training=training)
x = nn.gelu(x)
x = einn.Dropout("[...]", drop_rate=0.2)(x, training=training)
x = einn.Linear("b [...->c]", c=10)(x)
return x
net = Net()
inputs, labels = next(iter(trainloader))
params = net.init(
{"dropout": next_rng(), "params": next_rng()}, jnp.asarray(inputs), training=True
) # Run on dummy batch
if "stats" not in params:
params["stats"] = {}
optimizer = optax.adam(3e-4)
opt_state = optimizer.init(params["params"])
@partial(jax.jit, donate_argnums=(0, 1))
def update_step(params, opt_state, images, labels, rng):
def loss_fn(params, stats):
logits, new_stats = net.apply(
{"params": params, "stats": stats},
images,
training=True,
rngs={"dropout": rng},
mutable=["stats"],
)
one_hot = jax.nn.one_hot(labels, 10)
loss = jnp.mean(optax.softmax_cross_entropy(logits=logits, labels=one_hot))
return loss, new_stats
(_loss, new_stats), grads = jax.value_and_grad(loss_fn, has_aux=True)(
params["params"], params["stats"]
)
params["stats"] = new_stats["stats"]
updates, new_opt_state = optimizer.update(grads, opt_state, params["params"])
params["params"] = optax.apply_updates(params["params"], updates)
return params, new_opt_state
@jax.jit
def test_step(params, images, labels):
logits = net.apply(params, images, training=False)
accurate = jnp.argmax(logits, axis=1) == jnp.asarray(labels)
return accurate
print("Starting training")
for epoch in range(100):
t0 = time.time()
# Train
for data in trainloader:
inputs, labels = data
params, opt_state = update_step(
params, opt_state, jnp.asarray(inputs), jnp.asarray(labels), next_rng()
)
# Test
correct = 0
total = 0
for data in testloader:
images, labels = data
accurate = test_step(params, jnp.asarray(images), jnp.asarray(labels))
total += accurate.shape[0]
correct += jnp.sum(accurate)
print(
f"Test accuracy after {epoch + 1:5d} epochs: {float(correct) / total} "
f"({time.time() - t0:.2f}sec)"
)
|
