-
Notifications
You must be signed in to change notification settings - Fork 8
/
train_synthetic.py
164 lines (145 loc) · 5.9 KB
/
train_synthetic.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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
"""Main training script for synthetic problems."""
import argparse
import os
import time
import scipy.stats as st
import wandb
import numpy as np
import torch
from torch import nn
from torch import optim
import torch.nn.functional as F
import higher
import layers
from synthetic_loader import SyntheticLoader
from inner_optimizers import InnerOptBuilder
OUTPUT_PATH = "./outputs/synthetic_outputs"
def train(step_idx, data, net, inner_opt_builder, meta_opt, n_inner_iter):
"""Main meta-training step."""
x_spt, y_spt, x_qry, y_qry = data
task_num = x_spt.size()[0]
querysz = x_qry.size(1)
inner_opt = inner_opt_builder.inner_opt
qry_losses = []
meta_opt.zero_grad()
for i in range(task_num):
with higher.innerloop_ctx(
net,
inner_opt,
copy_initial_weights=False,
override=inner_opt_builder.overrides,
) as (
fnet,
diffopt,
):
for _ in range(n_inner_iter):
spt_pred = fnet(x_spt[i])
spt_loss = F.mse_loss(spt_pred, y_spt[i])
diffopt.step(spt_loss)
qry_pred = fnet(x_qry[i])
qry_loss = F.mse_loss(qry_pred, y_qry[i])
qry_losses.append(qry_loss.detach().cpu().numpy())
qry_loss.backward()
metrics = {"train_loss": np.mean(qry_losses)}
wandb.log(metrics, step=step_idx)
meta_opt.step()
def test(step_idx, data, net, inner_opt_builder, n_inner_iter):
"""Main meta-training step."""
x_spt, y_spt, x_qry, y_qry = data
task_num = x_spt.size()[0]
querysz = x_qry.size(1)
inner_opt = inner_opt_builder.inner_opt
qry_losses = []
for i in range(task_num):
with higher.innerloop_ctx(
net, inner_opt, track_higher_grads=False, override=inner_opt_builder.overrides,
) as (
fnet,
diffopt,
):
for _ in range(n_inner_iter):
spt_pred = fnet(x_spt[i])
spt_loss = F.mse_loss(spt_pred, y_spt[i])
diffopt.step(spt_loss)
qry_pred = fnet(x_qry[i])
qry_loss = F.mse_loss(qry_pred, y_qry[i])
qry_losses.append(qry_loss.detach().cpu().numpy())
avg_qry_loss = np.mean(qry_losses)
_low, high = st.t.interval(
0.95, len(qry_losses) - 1, loc=avg_qry_loss, scale=st.sem(qry_losses)
)
test_metrics = {"test_loss": avg_qry_loss, "test_err": high - avg_qry_loss}
wandb.log(test_metrics, step=step_idx)
return avg_qry_loss
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--init_inner_lr", type=float, default=0.1)
parser.add_argument("--outer_lr", type=float, default=0.001)
parser.add_argument("--k_spt", type=int, default=1)
parser.add_argument("--k_qry", type=int, default=19)
parser.add_argument("--lr_mode", type=str, default="per_layer")
parser.add_argument("--num_inner_steps", type=int, default=1)
parser.add_argument("--num_outer_steps", type=int, default=1000)
parser.add_argument("--inner_opt", type=str, default="maml")
parser.add_argument("--outer_opt", type=str, default="Adam")
parser.add_argument("--problem", type=str, default="rank1")
parser.add_argument("--model", type=str, default="conv")
parser.add_argument("--device", type=str, default="cpu")
if not os.path.exists(OUTPUT_PATH):
os.makedirs(OUTPUT_PATH)
wandb.init(project="weight_sharing_toy", dir=OUTPUT_PATH)
args = parser.parse_args()
wandb.config.update(args)
cfg = wandb.config
device = torch.device(cfg.device)
db = SyntheticLoader(device, problem=cfg.problem, k_spt=cfg.k_spt, k_qry=cfg.k_qry)
if cfg.problem in ["2d_rot8_flip", "2d_rot8"]:
c_o = 24 if cfg.problem == "2d_rot8" else 48
if cfg.model == "share_conv":
net = nn.Sequential(layers.ShareConv2d(1, c_o, 3, bias=False)).to(device)
elif cfg.model == "conv":
net = nn.Sequential(nn.Conv2d(1, c_o, 3, bias=False)).to(device)
else:
raise ValueError(f"Invalid model {cfg.model}")
elif cfg.problem in ["rank1", "rank2", "rank5"]:
if cfg.model == "lc":
net = nn.Sequential(layers.LocallyConnected1d(1, 1, 68, kernel_size=3, bias=False)).to(
device
)
elif cfg.model == "fc":
net = nn.Sequential(nn.Linear(70, 68, bias=False)).to(device)
elif cfg.model == "conv":
net = nn.Sequential(nn.Conv1d(1, 1, kernel_size=3, bias=False)).to(device)
elif cfg.model == "share_fc":
latent = {"rank1": 3, "rank2": 6, "rank5": 30}[cfg.problem]
net = nn.Sequential(layers.ShareLinearFull(70, 68, bias=False, latent_size=latent)).to(
device
)
else:
raise ValueError(f"Invalid model {cfg.model}")
inner_opt_builder = InnerOptBuilder(
net, device, cfg.inner_opt, cfg.init_inner_lr, "learned", cfg.lr_mode
)
if cfg.outer_opt == "SGD":
meta_opt = optim.SGD(inner_opt_builder.metaparams.values(), lr=cfg.outer_lr)
else:
meta_opt = optim.Adam(inner_opt_builder.metaparams.values(), lr=cfg.outer_lr)
start_time = time.time()
for step_idx in range(cfg.num_outer_steps):
data, _filters = db.next(32, "train")
train(step_idx, data, net, inner_opt_builder, meta_opt, cfg.num_inner_steps)
if step_idx == 0 or (step_idx + 1) % 100 == 0:
test_data, _filters = db.next(300, "test")
val_loss = test(
step_idx,
test_data,
net,
inner_opt_builder,
cfg.num_inner_steps,
)
if step_idx > 0:
steps_p_sec = (step_idx + 1) / (time.time() - start_time)
wandb.log({"steps_per_sec": steps_p_sec}, step=step_idx)
print(f"Step: {step_idx}. Steps/sec: {steps_p_sec:.2f}")
if __name__ == "__main__":
main()