-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathfedavg_quantization.py
145 lines (124 loc) · 6.51 KB
/
fedavg_quantization.py
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
132
133
134
135
136
137
138
139
140
141
142
143
144
145
import chex
import jax
import jax.numpy as jnp # JAX NumPy
import numpy as np
import tensorflow_datasets as tfds # TFDS for MNIST
import wandb
from evosax import NetworkMapper
from backprop import sl
from args import get_args
from utils import helpers
from flax.core import FrozenDict
from evosax import ParameterReshaper
import os
os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true'
# compress the array with quantization based on array distribution
def quantize(array, min_val, max_val, n_bits):
# max_val = array.max()
# min_val = array.min()
step = (max_val - min_val) / (2 ** n_bits - 1)
array = ((array - min_val) / step).round()
return array
# dequantization array
def dequantize(array, min_val, max_val, n_bits):
step = (max_val - min_val) / (2 ** n_bits - 1)
array = array * step + min_val
return array
def sparsify(array, percentage):
original = array
array = jnp.abs(array.flatten())
array = jnp.sort(array)
threshold = array[int(len(array) * percentage)]
array = jnp.where(jnp.abs(original) < threshold, 0, original)
return array
# L2 distance
def l2(x, y):
return -1 * jnp.sqrt(jnp.sum((x - y) ** 2)) # / jnp.sqrt(jnp.sum(x ** 2))
class TaskManager:
def __init__(self, rng: chex.PRNGKey, args):
wandb.run.name = '{}-{}-{} b{} c{} s{} q{} -- {}' \
.format(args.dataset, args.algo,
args.dist,
args.batch_size, args.n_clients,
args.seed, args.quantize_bits, wandb.run.id)
wandb.run.save()
self.train_ds, self.test_ds = sl.get_fed_datasets(args.dataset, args.n_clients, 2, args.dist == 'IID')
rng = jax.random.PRNGKey(0)
rng, init_rng = jax.random.split(rng)
self.learning_rate = wandb.config.lr
self.momentum = wandb.config.momentum
network = NetworkMapper[wandb.config.network_name](**wandb.config.network_config)
self.state = sl.create_train_state(init_rng, network, self.learning_rate, self.momentum)
del init_rng # Must not be used anymore.
self.param_count = sum(x.size for x in jax.tree_leaves(self.state.params))
self.num_epochs = wandb.config.n_rounds
self.batch_size = wandb.config.batch_size
self.client_epoch = wandb.config.client_epoch
self.n_clients = args.n_clients
min_cut = 10000
# if args.dataset == 'mnist':
# min_cut = 5421
self.X = jnp.array([train['image'][:min_cut] for train in self.train_ds])
self.y = jnp.array([train['label'][:min_cut] for train in self.train_ds])
self.args = args
self.n_bits = args.quantize_bits
self.param_reshaper = ParameterReshaper(self.state.params, n_devices=1)
def run(self, rng: chex.PRNGKey):
for epoch in range(0, self.num_epochs + 1):
# Use a separate PRNG key to permute image data during shuffling
rng, input_rng = jax.random.split(rng)
# Run an optimization step over a training batch
# clients = [self.state for i in range(5)]
clients, loss, acc = jax.vmap(sl.train_epoch, in_axes=(None, 0, 0, None, None))(self.state,
self.X,
self.y,
self.batch_size, input_rng)
for c_epoch in range(self.client_epoch):
input_rng, c_rng = jax.random.split(input_rng)
clients, loss, acc = jax.vmap(sl.train_epoch, in_axes=(0, 0, 0, None, None))(clients,
self.X,
self.y,
self.batch_size, c_rng)
wandb.log({
'Epoch': epoch * self.client_epoch + c_epoch,
'Train Loss': loss.mean(),
'Train Accuracy': acc.mean(),
})
server = self.param_reshaper.network_to_flat(self.state.params)
target_server = jax.vmap(self.param_reshaper.network_to_flat)(clients.params)
target_server = (target_server - server)
min_val, max_val = jax.vmap(jnp.min)(target_server), jax.vmap(jnp.max)(target_server)
target_server = jax.vmap(sparsify, in_axes=(0, None))(target_server, self.args.percentage)
# target_server = jax.vmap(quantize, in_axes=(0, 0, 0, None))(target_server, min_val, max_val, self.n_bits)
# target_server = jax.vmap(dequantize, in_axes=(0, 0, 0, None))(target_server, min_val, max_val, self.n_bits)
target_server = jax.vmap(jnp.mean)(target_server.T)
# target_server = jax.vmap(quantize, in_axes=(0, None))(target_server, self.n_bits)
# target_server = dequantize(target_server, min_val.mean(), max_val.mean(), self.n_bits)
target_server = sparsify(target_server, self.args.percentage)
target_server = target_server + server
params = self.param_reshaper.reshape_single_net(target_server)
self.state = self.state.replace(params=FrozenDict(params))
rng, eval_rng = jax.random.split(rng)
test_loss, test_accuracy = sl.eval_model(params, self.test_ds, eval_rng)
remining_params = self.param_count * (1 - self.args.percentage)
wandb.log({
'Round': epoch,
'Test Loss': test_loss,
'Global Accuracy': test_accuracy,
# 'Communication': epoch * 2 * self.param_count / (32 / self.n_bits),
'Communication': epoch * 2 * remining_params * ((self.n_bits + np.log2(self.param_count))/ 32),
})
def run():
print(jax.devices())
args = get_args()
config = helpers.load_config(args.config)
wandb.init(project='evofed-publish', config=args)
wandb.config.update(config)
args = wandb.config
rng = jax.random.PRNGKey(args.seed)
rng, rng_init, rng_run = jax.random.split(rng, 3)
manager = TaskManager(rng_init, args)
manager.run(rng_run)
if __name__ == '__main__':
# wandb.agent('tdt4lz81', function=run, project='evofed', count=10)
run()