iicd/xautodl
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update LFNA

Browse Source
This commit is contained in:
D-X-Y 2021-05-15 16:01:40 +08:00
parent b81ef2dd74
commit 72f240bf0a
12 changed files with 128 additions and 1050 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

239
exps/LFNA/backup/lfna-fix-init.py
View File

@ -1,239 +0,0 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
# python exps/LFNA/lfna-fix-init.py --env_version v1 --hidden_dim 16
#####################################################
import sys, time, copy, torch, random, argparse
from tqdm import tqdm
from copy import deepcopy
from pathlib import Path
lib_dir = (Path(__file__).parent / ".." / ".." / "lib").resolve()
if str(lib_dir) not in sys.path:
sys.path.insert(0, str(lib_dir))
from procedures import prepare_seed, prepare_logger, save_checkpoint, copy_checkpoint
from log_utils import time_string
from log_utils import AverageMeter, convert_secs2time
from utils import split_str2indexes
from procedures.advanced_main import basic_train_fn, basic_eval_fn
from procedures.metric_utils import SaveMetric, MSEMetric, ComposeMetric
from datasets.synthetic_core import get_synthetic_env
from models.xcore import get_model
from xlayers import super_core
from lfna_utils import lfna_setup, train_model, TimeData
class LFNAmlp:
"""A LFNA meta-model that uses the MLP as delta-net."""
def __init__(self, obs_dim, hidden_sizes, act_name, criterion):
self.delta_net = super_core.SuperSequential(
super_core.SuperLinear(obs_dim, hidden_sizes[0]),
super_core.super_name2activation[act_name](),
super_core.SuperLinear(hidden_sizes[0], hidden_sizes[1]),
super_core.super_name2activation[act_name](),
super_core.SuperLinear(hidden_sizes[1], 1),
)
self.meta_optimizer = torch.optim.Adam(
self.delta_net.parameters(), lr=0.001, amsgrad=True
)
self.criterion = criterion
def adapt(self, model, seq_datasets):
delta_inputs = []
container = model.get_w_container()
for iseq, dataset in enumerate(seq_datasets):
y_hat = model.forward_with_container(dataset.x, container)
loss = self.criterion(y_hat, dataset.y)
gradients = torch.autograd.grad(loss, container.parameters())
with torch.no_grad():
flatten_g = container.flatten(gradients)
delta_inputs.append(flatten_g)
flatten_w = container.no_grad_clone().flatten()
delta_inputs.append(flatten_w)
delta_inputs = torch.stack(delta_inputs, dim=-1)
delta = self.delta_net(delta_inputs)
delta = torch.clamp(delta, -0.8, 0.8)
unflatten_delta = container.unflatten(delta)
future_container = container.no_grad_clone().additive(unflatten_delta)
return future_container
def step(self):
torch.nn.utils.clip_grad_norm_(self.delta_net.parameters(), 1.0)
self.meta_optimizer.step()
def zero_grad(self):
self.meta_optimizer.zero_grad()
self.delta_net.zero_grad()
def state_dict(self):
return dict(
delta_net=self.delta_net.state_dict(),
meta_optimizer=self.meta_optimizer.state_dict(),
)
def main(args):
logger, env_info, model_kwargs = lfna_setup(args)
dynamic_env = env_info["dynamic_env"]
model = get_model(dict(model_type="simple_mlp"), **model_kwargs)
total_time = env_info["total"]
for i in range(total_time):
for xkey in ("timestamp", "x", "y"):
nkey = "{:}-{:}".format(i, xkey)
assert nkey in env_info, "{:} no in {:}".format(nkey, list(env_info.keys()))
train_time_bar = total_time // 2
network = get_model(dict(model_type="simple_mlp"), **model_kwargs)
criterion = torch.nn.MSELoss()
logger.log("There are {:} weights.".format(network.get_w_container().numel()))
adaptor = LFNAmlp(1 + args.meta_seq, (20, 20), "leaky_relu", criterion)
# pre-train the model
init_dataset = TimeData(0, env_info["0-x"], env_info["0-y"])
init_loss = train_model(network, init_dataset, args.init_lr, args.epochs)
logger.log("The pre-training loss is {:.4f}".format(init_loss))
# LFNA meta-training
meta_loss_meter = AverageMeter()
per_epoch_time, start_time = AverageMeter(), time.time()
for iepoch in range(args.epochs):
need_time = "Time Left: {:}".format(
convert_secs2time(per_epoch_time.avg * (args.epochs - iepoch), True)
)
logger.log(
"[{:}] [{:04d}/{:04d}] ".format(time_string(), iepoch, args.epochs)
+ need_time
)
adaptor.zero_grad()
batch_indexes, meta_losses = [], []
for ibatch in range(args.meta_batch):
sampled_timestamp = random.random() * train_time_bar
batch_indexes.append("{:.3f}".format(sampled_timestamp))
seq_datasets = []
for iseq in range(args.meta_seq + 1):
cur_time = sampled_timestamp + iseq * dynamic_env.timestamp_interval
cur_time, (x, y) = dynamic_env(cur_time)
seq_datasets.append(TimeData(cur_time, x, y))
history_datasets, future_dataset = seq_datasets[:-1], seq_datasets[-1]
future_container = adaptor.adapt(network, history_datasets)
future_y_hat = network.forward_with_container(
future_dataset.x, future_container
)
future_loss = adaptor.criterion(future_y_hat, future_dataset.y)
meta_losses.append(future_loss)
meta_loss = torch.stack(meta_losses).mean()
meta_loss.backward()
adaptor.step()
meta_loss_meter.update(meta_loss.item())
logger.log(
"meta-loss: {:.4f} ({:.4f}) batch: {:}".format(
meta_loss_meter.avg, meta_loss_meter.val, ",".join(batch_indexes[:5])
)
)
if iepoch % 200 == 0:
save_checkpoint(
{"adaptor": adaptor.state_dict(), "iepoch": iepoch},
logger.path("model"),
logger,
)
per_epoch_time.update(time.time() - start_time)
start_time = time.time()
w_container_per_epoch = dict()
for idx in range(1, env_info["total"]):
future_time = env_info["{:}-timestamp".format(idx)]
future_x = env_info["{:}-x".format(idx)]
future_y = env_info["{:}-y".format(idx)]
seq_datasets = []
for iseq in range(1, args.meta_seq + 1):
cur_time = future_time - iseq * dynamic_env.timestamp_interval
cur_time, (x, y) = dynamic_env(cur_time)
seq_datasets.append(TimeData(cur_time, x, y))
seq_datasets.reverse()
future_container = adaptor.adapt(network, seq_datasets)
w_container_per_epoch[idx] = future_container.no_grad_clone()
with torch.no_grad():
future_y_hat = network.forward_with_container(
future_x, w_container_per_epoch[idx]
)
future_loss = adaptor.criterion(future_y_hat, future_y)
logger.log("meta-test: [{:03d}] -> loss={:.4f}".format(idx, future_loss.item()))
save_checkpoint(
{"w_container_per_epoch": w_container_per_epoch},
logger.path(None) / "final-ckp.pth",
logger,
)
logger.log("-" * 200 + "\n")
logger.close()
if __name__ == "__main__":
parser = argparse.ArgumentParser("Use the data in the past.")
parser.add_argument(
"--save_dir",
type=str,
default="./outputs/lfna-synthetic/lfna-fix-init",
help="The checkpoint directory.",
)
parser.add_argument(
"--env_version",
type=str,
required=True,
help="The synthetic enviornment version.",
)
parser.add_argument(
"--hidden_dim",
type=int,
required=True,
help="The hidden dimension.",
)
#####
parser.add_argument(
"--init_lr",
type=float,
default=0.1,
help="The initial learning rate for the optimizer (default is Adam)",
)
parser.add_argument(
"--meta_batch",
type=int,
default=32,
help="The batch size for the meta-model",
)
parser.add_argument(
"--meta_seq",
type=int,
default=10,
help="The length of the sequence for meta-model.",
)
parser.add_argument(
"--epochs",
type=int,
default=1000,
help="The total number of epochs.",
)
# Random Seed
parser.add_argument("--rand_seed", type=int, default=-1, help="manual seed")
args = parser.parse_args()
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
args.save_dir = "{:}-{:}-d{:}".format(
args.save_dir, args.env_version, args.hidden_dim
)
main(args)

239
exps/LFNA/backup/lfna-test-hpnet.py
View File

@ -1,239 +0,0 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
# python exps/LFNA/lfna-test-hpnet.py --env_version v1 --hidden_dim 16 --layer_dim 16 --epochs 10000 --init_lr 0.01
# python exps/LFNA/lfna-test-hpnet.py --env_version v1 --hidden_dim 16 --layer_dim 16 --epochs 10000 --init_lr 0.01 --device cuda
#####################################################
import sys, time, copy, torch, random, argparse
from tqdm import tqdm
from copy import deepcopy
from pathlib import Path
lib_dir = (Path(__file__).parent / ".." / ".." / "lib").resolve()
if str(lib_dir) not in sys.path:
sys.path.insert(0, str(lib_dir))
from procedures import prepare_seed, prepare_logger, save_checkpoint, copy_checkpoint
from log_utils import time_string
from log_utils import AverageMeter, convert_secs2time
from utils import split_str2indexes
from procedures.advanced_main import basic_train_fn, basic_eval_fn
from procedures.metric_utils import SaveMetric, MSEMetric, ComposeMetric
from datasets.synthetic_core import get_synthetic_env
from models.xcore import get_model
from xlayers import super_core, trunc_normal_
from lfna_utils import lfna_setup, train_model, TimeData
# from lfna_models import HyperNet_VX as HyperNet
from lfna_models import HyperNet
def main(args):
logger, env_info, model_kwargs = lfna_setup(args)
dynamic_env = env_info["dynamic_env"]
model = get_model(**model_kwargs)
model = model.to(args.device)
criterion = torch.nn.MSELoss()
shape_container = model.get_w_container().to_shape_container()
total_bar = 100
hypernet = HyperNet(shape_container, args.layer_dim, args.task_dim, total_bar)
hypernet = hypernet.to(args.device)
logger.log(
"{:} There are {:} weights in the base-model.".format(
time_string(), model.numel()
)
)
logger.log(
"{:} There are {:} weights in the meta-model.".format(
time_string(), hypernet.numel()
)
)
for i in range(total_bar):
env_info["{:}-x".format(i)] = env_info["{:}-x".format(i)].to(args.device)
env_info["{:}-y".format(i)] = env_info["{:}-y".format(i)].to(args.device)
model.train()
hypernet.train()
optimizer = torch.optim.Adam(
hypernet.parameters(), lr=args.init_lr, weight_decay=1e-5, amsgrad=True
)
lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(
optimizer,
milestones=[
int(args.epochs * 0.8),
int(args.epochs * 0.9),
],
gamma=0.1,
)
# total_bar = env_info["total"] - 1
# LFNA meta-training
loss_meter = AverageMeter()
per_epoch_time, start_time = AverageMeter(), time.time()
last_success = 0
for iepoch in range(args.epochs):
need_time = "Time Left: {:}".format(
convert_secs2time(per_epoch_time.avg * (args.epochs - iepoch), True)
)
head_str = (
"[{:}] [{:04d}/{:04d}] ".format(time_string(), iepoch, args.epochs)
+ need_time
)
losses = []
# for ibatch in range(args.meta_batch):
for cur_time in range(total_bar):
# cur_time = random.randint(0, total_bar)
# cur_task_embed = task_embeds[cur_time]
cur_container = hypernet(cur_time)
cur_x = env_info["{:}-x".format(cur_time)]
cur_y = env_info["{:}-y".format(cur_time)]
cur_dataset = TimeData(cur_time, cur_x, cur_y)
preds = model.forward_with_container(cur_dataset.x, cur_container)
optimizer.zero_grad()
loss = criterion(preds, cur_dataset.y)
losses.append(loss)
final_loss = torch.stack(losses).mean()
final_loss.backward()
optimizer.step()
lr_scheduler.step()
loss_meter.update(final_loss.item())
success, best_score = hypernet.save_best(-loss_meter.val)
if success:
logger.log("Achieve the best with best_score = {:.3f}".format(best_score))
last_success = iepoch
if iepoch - last_success >= args.early_stop_thresh:
logger.log("Early stop at {:}".format(iepoch))
break
if iepoch % 20 == 0:
logger.log(
head_str
+ " meta-loss: {:.4f} ({:.4f}) :: lr={:.5f}, batch={:}".format(
loss_meter.avg,
loss_meter.val,
min(lr_scheduler.get_last_lr()),
len(losses),
)
)
save_checkpoint(
{
"hypernet": hypernet.state_dict(),
"lr_scheduler": lr_scheduler.state_dict(),
"iepoch": iepoch,
},
logger.path("model"),
logger,
)
loss_meter.reset()
per_epoch_time.update(time.time() - start_time)
start_time = time.time()
print(model)
print(hypernet)
hypernet.load_best()
w_container_per_epoch = dict()
for idx in range(0, total_bar):
future_time = env_info["{:}-timestamp".format(idx)]
future_x = env_info["{:}-x".format(idx)]
future_y = env_info["{:}-y".format(idx)]
# future_container = hypernet(task_embeds[idx])
future_container = hypernet(idx)
w_container_per_epoch[idx] = future_container.no_grad_clone()
with torch.no_grad():
future_y_hat = model.forward_with_container(
future_x, w_container_per_epoch[idx]
)
future_loss = criterion(future_y_hat, future_y)
logger.log("meta-test: [{:03d}] -> loss={:.4f}".format(idx, future_loss.item()))
save_checkpoint(
{"w_container_per_epoch": w_container_per_epoch},
logger.path(None) / "final-ckp.pth",
logger,
)
logger.log("-" * 200 + "\n")
logger.close()
if __name__ == "__main__":
parser = argparse.ArgumentParser("Use the data in the past.")
parser.add_argument(
"--save_dir",
type=str,
default="./outputs/lfna-synthetic/lfna-test-hpnet",
help="The checkpoint directory.",
)
parser.add_argument(
"--env_version",
type=str,
required=True,
help="The synthetic enviornment version.",
)
parser.add_argument(
"--hidden_dim",
type=int,
required=True,
help="The hidden dimension.",
)
parser.add_argument(
"--layer_dim",
type=int,
required=True,
help="The hidden dimension.",
)
parser.add_argument(
"--early_stop_thresh",
type=int,
default=100,
help="The maximum epochs for early stop.",
)
#####
parser.add_argument(
"--init_lr",
type=float,
default=0.1,
help="The initial learning rate for the optimizer (default is Adam)",
)
parser.add_argument(
"--meta_batch",
type=int,
default=64,
help="The batch size for the meta-model",
)
parser.add_argument(
"--epochs",
type=int,
default=2000,
help="The total number of epochs.",
)
parser.add_argument(
"--device",
type=str,
default="cpu",
help="",
)
# Random Seed
parser.add_argument("--rand_seed", type=int, default=-1, help="manual seed")
args = parser.parse_args()
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
args.task_dim = args.layer_dim
args.save_dir = "{:}-{:}-d{:}".format(
args.save_dir, args.env_version, args.hidden_dim
)
main(args)

134
exps/LFNA/backup/lfna-ttss-hpnet.py
View File

@ -1,134 +0,0 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
# python exps/LFNA/lfna-ttss-hpnet.py --env_version v1 --hidden_dim 16
#####################################################
import sys, time, copy, torch, random, argparse
from tqdm import tqdm
from copy import deepcopy
from pathlib import Path
lib_dir = (Path(__file__).parent / ".." / ".." / "lib").resolve()
if str(lib_dir) not in sys.path:
sys.path.insert(0, str(lib_dir))
from procedures import prepare_seed, prepare_logger, save_checkpoint, copy_checkpoint
from log_utils import time_string
from log_utils import AverageMeter, convert_secs2time
from utils import split_str2indexes
from procedures.advanced_main import basic_train_fn, basic_eval_fn
from procedures.metric_utils import SaveMetric, MSEMetric, ComposeMetric
from datasets.synthetic_core import get_synthetic_env
from models.xcore import get_model
from xlayers import super_core
from lfna_utils import lfna_setup, train_model, TimeData
from lfna_models import HyperNet_VX as HyperNet
def main(args):
logger, env_info, model_kwargs = lfna_setup(args)
dynamic_env = env_info["dynamic_env"]
model = get_model(**model_kwargs)
total_time = env_info["total"]
for i in range(total_time):
for xkey in ("timestamp", "x", "y"):
nkey = "{:}-{:}".format(i, xkey)
assert nkey in env_info, "{:} no in {:}".format(nkey, list(env_info.keys()))
train_time_bar = total_time // 2
criterion = torch.nn.MSELoss()
logger.log("There are {:} weights.".format(model.get_w_container().numel()))
# pre-train the model
dataset = init_dataset = TimeData(0, env_info["0-x"], env_info["0-y"])
shape_container = model.get_w_container().to_shape_container()
hypernet = HyperNet(shape_container, 16)
print(hypernet)
optimizer = torch.optim.Adam(hypernet.parameters(), lr=args.init_lr, amsgrad=True)
best_loss, best_param = None, None
for _iepoch in range(args.epochs):
container = hypernet(None)
preds = model.forward_with_container(dataset.x, container)
optimizer.zero_grad()
loss = criterion(preds, dataset.y)
loss.backward()
optimizer.step()
# save best
if best_loss is None or best_loss > loss.item():
best_loss = loss.item()
best_param = copy.deepcopy(model.state_dict())
print("hyper-net : best={:.4f}".format(best_loss))
init_loss = train_model(model, init_dataset, args.init_lr, args.epochs)
logger.log("The pre-training loss is {:.4f}".format(init_loss))
print(model)
print(hypernet)
logger.log("-" * 200 + "\n")
logger.close()
if __name__ == "__main__":
parser = argparse.ArgumentParser("Use the data in the past.")
parser.add_argument(
"--save_dir",
type=str,
default="./outputs/lfna-synthetic/lfna-debug",
help="The checkpoint directory.",
)
parser.add_argument(
"--env_version",
type=str,
required=True,
help="The synthetic enviornment version.",
)
parser.add_argument(
"--hidden_dim",
type=int,
required=True,
help="The hidden dimension.",
)
#####
parser.add_argument(
"--init_lr",
type=float,
default=0.1,
help="The initial learning rate for the optimizer (default is Adam)",
)
parser.add_argument(
"--meta_batch",
type=int,
default=32,
help="The batch size for the meta-model",
)
parser.add_argument(
"--meta_seq",
type=int,
default=10,
help="The length of the sequence for meta-model.",
)
parser.add_argument(
"--epochs",
type=int,
default=2000,
help="The total number of epochs.",
)
# Random Seed
parser.add_argument("--rand_seed", type=int, default=-1, help="manual seed")
args = parser.parse_args()
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
args.save_dir = "{:}-{:}-d{:}".format(
args.save_dir, args.env_version, args.hidden_dim
)
main(args)

272
exps/LFNA/backup/lfna-v1.py
View File

@ -1,272 +0,0 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
# python exps/LFNA/lfna-v1.py
#####################################################
import sys, time, copy, torch, random, argparse
from tqdm import tqdm
from copy import deepcopy
from pathlib import Path
lib_dir = (Path(__file__).parent / ".." / ".." / "lib").resolve()
if str(lib_dir) not in sys.path:
sys.path.insert(0, str(lib_dir))
from procedures import prepare_seed, prepare_logger, save_checkpoint, copy_checkpoint
from log_utils import time_string
from log_utils import AverageMeter, convert_secs2time
from utils import split_str2indexes
from procedures.advanced_main import basic_train_fn, basic_eval_fn
from procedures.metric_utils import SaveMetric, MSEMetric, ComposeMetric
from datasets.synthetic_core import get_synthetic_env
from models.xcore import get_model
from xlayers import super_core
class LFNAmlp:
"""A LFNA meta-model that uses the MLP as delta-net."""
def __init__(self, obs_dim, hidden_sizes, act_name):
self.delta_net = super_core.SuperSequential(
super_core.SuperLinear(obs_dim, hidden_sizes[0]),
super_core.super_name2activation[act_name](),
super_core.SuperLinear(hidden_sizes[0], hidden_sizes[1]),
super_core.super_name2activation[act_name](),
super_core.SuperLinear(hidden_sizes[1], 1),
)
self.meta_optimizer = torch.optim.Adam(
self.delta_net.parameters(), lr=0.01, amsgrad=True
)
def adapt(self, model, criterion, w_container, seq_datasets):
w_container.requires_grad_(True)
containers = [w_container]
for idx, dataset in enumerate(seq_datasets):
x, y = dataset.x, dataset.y
y_hat = model.forward_with_container(x, containers[-1])
loss = criterion(y_hat, y)
gradients = torch.autograd.grad(loss, containers[-1].tensors)
with torch.no_grad():
flatten_w = containers[-1].flatten().view(-1, 1)
flatten_g = containers[-1].flatten(gradients).view(-1, 1)
input_statistics = torch.tensor([x.mean(), x.std()]).view(1, 2)
input_statistics = input_statistics.expand(flatten_w.numel(), -1)
delta_inputs = torch.cat((flatten_w, flatten_g, input_statistics), dim=-1)
delta = self.delta_net(delta_inputs).view(-1)
delta = torch.clamp(delta, -0.5, 0.5)
unflatten_delta = containers[-1].unflatten(delta)
future_container = containers[-1].no_grad_clone().additive(unflatten_delta)
# future_container = containers[-1].additive(unflatten_delta)
containers.append(future_container)
# containers = containers[1:]
meta_loss = []
temp_containers = []
for idx, dataset in enumerate(seq_datasets):
if idx == 0:
continue
current_container = containers[idx]
y_hat = model.forward_with_container(dataset.x, current_container)
loss = criterion(y_hat, dataset.y)
meta_loss.append(loss)
temp_containers.append((dataset.timestamp, current_container, -loss.item()))
meta_loss = sum(meta_loss)
w_container.requires_grad_(False)
# meta_loss.backward()
# self.meta_optimizer.step()
return meta_loss, temp_containers
def step(self):
torch.nn.utils.clip_grad_norm_(self.delta_net.parameters(), 1.0)
self.meta_optimizer.step()
def zero_grad(self):
self.meta_optimizer.zero_grad()
self.delta_net.zero_grad()
class TimeData:
def __init__(self, timestamp, xs, ys):
self._timestamp = timestamp
self._xs = xs
self._ys = ys
@property
def x(self):
return self._xs
@property
def y(self):
return self._ys
@property
def timestamp(self):
return self._timestamp
class Population:
"""A population used to maintain models at different timestamps."""
def __init__(self):
self._time2model = dict()
self._time2score = dict() # higher is better
def append(self, timestamp, model, score):
if timestamp in self._time2model:
if self._time2score[timestamp] > score:
return
self._time2model[timestamp] = model.no_grad_clone()
self._time2score[timestamp] = score
def query(self, timestamp):
closet_timestamp = None
for xtime, model in self._time2model.items():
if closet_timestamp is None or (
xtime < timestamp and timestamp - closet_timestamp >= timestamp - xtime
):
closet_timestamp = xtime
return self._time2model[closet_timestamp], closet_timestamp
def debug_info(self, timestamps):
xstrs = []
for timestamp in timestamps:
if timestamp in self._time2score:
xstrs.append(
"{:04d}: {:.4f}".format(timestamp, self._time2score[timestamp])
)
return ", ".join(xstrs)
def main(args):
prepare_seed(args.rand_seed)
logger = prepare_logger(args)
cache_path = (logger.path(None) / ".." / "env-info.pth").resolve()
if cache_path.exists():
env_info = torch.load(cache_path)
else:
env_info = dict()
dynamic_env = get_synthetic_env()
env_info["total"] = len(dynamic_env)
for idx, (timestamp, (_allx, _ally)) in enumerate(tqdm(dynamic_env)):
env_info["{:}-timestamp".format(idx)] = timestamp
env_info["{:}-x".format(idx)] = _allx
env_info["{:}-y".format(idx)] = _ally
env_info["dynamic_env"] = dynamic_env
torch.save(env_info, cache_path)
total_time = env_info["total"]
for i in range(total_time):
for xkey in ("timestamp", "x", "y"):
nkey = "{:}-{:}".format(i, xkey)
assert nkey in env_info, "{:} no in {:}".format(nkey, list(env_info.keys()))
train_time_bar = total_time // 2
base_model = get_model(
dict(model_type="simple_mlp"),
act_cls="leaky_relu",
norm_cls="identity",
input_dim=1,
output_dim=1,
)
w_container = base_model.get_w_container()
criterion = torch.nn.MSELoss()
print("There are {:} weights.".format(w_container.numel()))
adaptor = LFNAmlp(4, (50, 20), "leaky_relu")
pool = Population()
pool.append(0, w_container, -100)
# LFNA meta-training
per_epoch_time, start_time = AverageMeter(), time.time()
for iepoch in range(args.epochs):
need_time = "Time Left: {:}".format(
convert_secs2time(per_epoch_time.avg * (args.epochs - iepoch), True)
)
logger.log(
"[{:}] [{:04d}/{:04d}] ".format(time_string(), iepoch, args.epochs)
+ need_time
)
adaptor.zero_grad()
debug_timestamp = set()
all_meta_losses = []
for ibatch in range(args.meta_batch):
sampled_timestamp = random.randint(0, train_time_bar)
query_w_container, query_timestamp = pool.query(sampled_timestamp)
# def adapt(self, model, w_container, xs, ys):
seq_datasets = []
# xs, ys = [], []
for it in range(sampled_timestamp, sampled_timestamp + args.max_seq):
xs = env_info["{:}-x".format(it)]
ys = env_info["{:}-y".format(it)]
seq_datasets.append(TimeData(it, xs, ys))
temp_meta_loss, temp_containers = adaptor.adapt(
base_model, criterion, query_w_container, seq_datasets
)
all_meta_losses.append(temp_meta_loss)
for temp_time, temp_container, temp_score in temp_containers:
pool.append(temp_time, temp_container, temp_score)
debug_timestamp.add(temp_time)
meta_loss = torch.stack(all_meta_losses).mean()
meta_loss.backward()
adaptor.step()
debug_str = pool.debug_info(debug_timestamp)
logger.log("meta-loss: {:.4f}".format(meta_loss.item()))
per_epoch_time.update(time.time() - start_time)
start_time = time.time()
logger.log("-" * 200 + "\n")
logger.close()
if __name__ == "__main__":
parser = argparse.ArgumentParser("Use the data in the past.")
parser.add_argument(
"--save_dir",
type=str,
default="./outputs/lfna-synthetic/lfna-v1",
help="The checkpoint directory.",
)
parser.add_argument(
"--init_lr",
type=float,
default=0.1,
help="The initial learning rate for the optimizer (default is Adam)",
)
parser.add_argument(
"--meta_batch",
type=int,
default=5,
help="The batch size for the meta-model",
)
parser.add_argument(
"--epochs",
type=int,
default=1000,
help="The total number of epochs.",
)
parser.add_argument(
"--max_seq",
type=int,
default=5,
help="The maximum length of the sequence.",
)
parser.add_argument(
"--workers",
type=int,
default=4,
help="The number of data loading workers (default: 4)",
)
# Random Seed
parser.add_argument("--rand_seed", type=int, default=-1, help="manual seed")
args = parser.parse_args()
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
main(args)

50
exps/LFNA/backup/lfna_models.py
View File

@ -1,50 +0,0 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
import copy
import torch
from xlayers import super_core
from xlayers import trunc_normal_
from models.xcore import get_model
class HyperNet(super_core.SuperModule):
def __init__(self, shape_container, input_embeding, return_container=True):
super(HyperNet, self).__init__()
self._shape_container = shape_container
self._num_layers = len(shape_container)
self._numel_per_layer = []
for ilayer in range(self._num_layers):
self._numel_per_layer.append(shape_container[ilayer].numel())
self.register_parameter(
"_super_layer_embed",
torch.nn.Parameter(torch.Tensor(self._num_layers, input_embeding)),
)
trunc_normal_(self._super_layer_embed, std=0.02)
model_kwargs = dict(
input_dim=input_embeding,
output_dim=max(self._numel_per_layer),
hidden_dim=input_embeding * 4,
act_cls="sigmoid",
norm_cls="identity",
)
self._generator = get_model(dict(model_type="simple_mlp"), **model_kwargs)
self._return_container = return_container
print("generator: {:}".format(self._generator))
def forward_raw(self, input):
weights = self._generator(self._super_layer_embed)
if self._return_container:
weights = torch.split(weights, 1)
return self._shape_container.translate(weights)
else:
return weights
def forward_candidate(self, input):
raise NotImplementedError
def extra_repr(self) -> str:
return "(_super_layer_embed): {:}".format(list(self._super_layer_embed.shape))

173
exps/LFNA/basic-maml.py
View File

@ -1,7 +1,7 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
# python exps/LFNA/basic-maml.py --env_version v1 --hidden_dim 16 --inner_step 5
# python exps/LFNA/basic-maml.py --env_version v1 --inner_step 5
# python exps/LFNA/basic-maml.py --env_version v2
#####################################################
import sys, time, copy, torch, random, argparse
@ -20,7 +20,7 @@ from utils import split_str2indexes
from procedures.advanced_main import basic_train_fn, basic_eval_fn
from procedures.metric_utils import SaveMetric, MSEMetric, ComposeMetric
from datasets.synthetic_core import get_synthetic_env
from datasets.synthetic_core import get_synthetic_env, EnvSampler
from models.xcore import get_model
from xlayers import super_core
@ -42,11 +42,10 @@ class MAML:
self.meta_lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(
self.meta_optimizer,
milestones=[
int(epochs * 0.25),
int(epochs * 0.5),
int(epochs * 0.75),
int(epochs * 0.8),
int(epochs * 0.9),
],
gamma=0.3,
gamma=0.1,
)
self.inner_lr = inner_lr
self.inner_step = inner_step
@ -85,33 +84,27 @@ class MAML:
self.meta_optimizer.load_state_dict(state_dict["meta_optimizer"])
self.meta_lr_scheduler.load_state_dict(state_dict["meta_lr_scheduler"])
def save_best(self, iepoch, score):
if self._best_info["score"] is None or self._best_info["score"] < score:
state_dict = dict(
criterion=self.criterion.state_dict(),
network=self.network.state_dict(),
meta_optimizer=self.meta_optimizer.state_dict(),
meta_lr_scheduler=self.meta_lr_scheduler.state_dict(),
)
self._best_info["state_dict"] = state_dict
self._best_info["score"] = score
self._best_info["iepoch"] = iepoch
is_best = True
else:
is_best = False
return self._best_info, is_best
def state_dict(self):
state_dict = dict()
state_dict["criterion"] = self.criterion.state_dict()
state_dict["network"] = self.network.state_dict()
state_dict["meta_optimizer"] = self.meta_optimizer.state_dict()
state_dict["meta_lr_scheduler"] = self.meta_lr_scheduler.state_dict()
return state_dict
def save_best(self, score):
success, best_score = self.network.save_best(score)
return success, best_score
def load_best(self):
self.network.load_best()
def main(args):
logger, env_info, model_kwargs = lfna_setup(args)
model = get_model(dict(model_type="simple_mlp"), **model_kwargs)
model = get_model(**model_kwargs)
total_time = env_info["total"]
for i in range(total_time):
for xkey in ("timestamp", "x", "y"):
nkey = "{:}-{:}".format(i, xkey)
assert nkey in env_info, "{:} no in {:}".format(nkey, list(env_info.keys()))
train_time_bar = total_time // 2
dynamic_env = get_synthetic_env(mode="train", version=args.env_version)
criterion = torch.nn.MSELoss()
@ -120,83 +113,65 @@ def main(args):
)
# meta-training
last_success_epoch = 0
per_epoch_time, start_time = AverageMeter(), time.time()
# for iepoch in range(args.epochs):
iepoch = 0
while iepoch < args.epochs:
for iepoch in range(args.epochs):
need_time = "Time Left: {:}".format(
convert_secs2time(per_epoch_time.avg * (args.epochs - iepoch), True)
)
logger.log(
head_str = (
"[{:}] [{:04d}/{:04d}] ".format(time_string(), iepoch, args.epochs)
+ need_time
)
maml.zero_grad()
batch_indexes, meta_losses = [], []
meta_losses = []
for ibatch in range(args.meta_batch):
sampled_timestamp = random.randint(0, train_time_bar)
batch_indexes.append("{:5d}".format(sampled_timestamp))
past_dataset = TimeData(
sampled_timestamp,
env_info["{:}-x".format(sampled_timestamp)],
env_info["{:}-y".format(sampled_timestamp)],
future_timestamp = dynamic_env.random_timestamp()
_, (future_x, future_y) = dynamic_env(future_timestamp)
past_timestamp = (
future_timestamp - args.prev_time * dynamic_env.timestamp_interval
)
future_dataset = TimeData(
sampled_timestamp + 1,
env_info["{:}-x".format(sampled_timestamp + 1)],
env_info["{:}-y".format(sampled_timestamp + 1)],
)
future_container = maml.adapt(past_dataset)
future_y_hat = maml.predict(future_dataset.x, future_container)
future_loss = maml.criterion(future_y_hat, future_dataset.y)
_, (past_x, past_y) = dynamic_env(past_timestamp)
future_container = maml.adapt(TimeData(past_timestamp, past_x, past_y))
future_y_hat = maml.predict(future_x, future_container)
future_loss = maml.criterion(future_y_hat, future_y)
meta_losses.append(future_loss)
meta_loss = torch.stack(meta_losses).mean()
meta_loss.backward()
maml.step()
logger.log(
"meta-loss: {:.4f} batch: {:}".format(
meta_loss.item(), ",".join(batch_indexes)
)
)
best_info, is_best = maml.save_best(iepoch, -meta_loss.item())
if is_best:
save_checkpoint(best_info, logger.path("best"), logger)
logger.log("Save the best into {:}".format(logger.path("best")))
if iepoch >= 10 and (
torch.isnan(meta_loss).item() or meta_loss.item() >= args.fail_thresh
):
xdata = torch.load(logger.path("best"))
maml.load_state_dict(xdata["state_dict"])
iepoch = xdata["iepoch"]
logger.log(
"The training failed, re-use the previous best epoch [{:}]".format(
iepoch
)
)
else:
iepoch = iepoch + 1
logger.log(head_str + " meta-loss: {:.4f}".format(meta_loss.item()))
success, best_score = maml.save_best(-meta_loss.item())
if success:
logger.log("Achieve the best with best_score = {:.3f}".format(best_score))
save_checkpoint(maml.state_dict(), logger.path("model"), logger)
last_success_epoch = iepoch
if iepoch - last_success_epoch >= args.early_stop_thresh:
logger.log("Early stop at {:}".format(iepoch))
break
per_epoch_time.update(time.time() - start_time)
start_time = time.time()
# meta-test
maml.load_best()
eval_env = env_info["dynamic_env"]
assert eval_env.timestamp_interval == dynamic_env.timestamp_interval
w_container_per_epoch = dict()
for idx in range(1, env_info["total"]):
past_dataset = TimeData(
idx - 1,
env_info["{:}-x".format(idx - 1)],
env_info["{:}-y".format(idx - 1)],
for idx in range(args.prev_time, len(eval_env)):
future_timestamp, (future_x, future_y) = eval_env[idx]
past_timestamp = (
future_timestamp.item() - args.prev_time * eval_env.timestamp_interval
)
current_container = maml.adapt(past_dataset)
w_container_per_epoch[idx] = current_container.no_grad_clone()
_, (past_x, past_y) = eval_env(past_timestamp)
future_container = maml.adapt(TimeData(past_timestamp, past_x, past_y))
w_container_per_epoch[idx] = future_container.no_grad_clone()
with torch.no_grad():
current_x = env_info["{:}-x".format(idx)]
current_y = env_info["{:}-y".format(idx)]
current_y_hat = maml.predict(current_x, w_container_per_epoch[idx])
current_loss = maml.criterion(current_y_hat, current_y)
logger.log(
"meta-test: [{:03d}] -> loss={:.4f}".format(idx, current_loss.item())
)
future_y_hat = maml.predict(future_x, w_container_per_epoch[idx])
future_loss = maml.criterion(future_y_hat, future_y)
logger.log("meta-test: [{:03d}] -> loss={:.4f}".format(idx, future_loss.item()))
save_checkpoint(
{"w_container_per_epoch": w_container_per_epoch},
logger.path(None) / "final-ckp.pth",
@ -224,13 +199,13 @@ if __name__ == "__main__":
parser.add_argument(
"--hidden_dim",
type=int,
required=True,
default=16,
help="The hidden dimension.",
)
parser.add_argument(
"--meta_lr",
type=float,
default=0.05,
default=0.01,
help="The learning rate for the MAML optimizer (default is Adam)",
)
parser.add_argument(
@ -242,24 +217,36 @@ if __name__ == "__main__":
parser.add_argument(
"--inner_lr",
type=float,
default=0.01,
default=0.005,
help="The learning rate for the inner optimization",
)
parser.add_argument(
"--inner_step", type=int, default=1, help="The inner loop steps for MAML."
)
parser.add_argument(
"--prev_time",
type=int,
default=5,
help="The gap between prev_time and current_timestamp",
)
parser.add_argument(
"--meta_batch",
type=int,
default=10,
default=64,
help="The batch size for the meta-model",
)
parser.add_argument(
"--epochs",
type=int,
default=1000,
default=2000,
help="The total number of epochs.",
)
parser.add_argument(
"--early_stop_thresh",
type=int,
default=50,
help="The maximum epochs for early stop.",
)
parser.add_argument(
"--workers",
type=int,
@ -272,7 +259,13 @@ if __name__ == "__main__":
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
args.save_dir = "{:}-s{:}-{:}-d{:}".format(
args.save_dir, args.inner_step, args.env_version, args.hidden_dim
args.save_dir = "{:}-s{:}-mlr{:}-d{:}-prev{:}-e{:}-env{:}".format(
args.save_dir,
args.inner_step,
args.meta_lr,
args.hidden_dim,
args.prev_time,
args.epochs,
args.env_version,
)
main(args)

23
exps/LFNA/basic-prev.py
View File

@ -1,7 +1,7 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
# python exps/LFNA/basic-prev.py --env_version v1 --hidden_dim 16 --epochs 500 --init_lr 0.1
# python exps/LFNA/basic-prev.py --env_version v1 --prev_time 5 --hidden_dim 16 --epochs 500 --init_lr 0.1
# python exps/LFNA/basic-prev.py --env_version v2 --hidden_dim 16 --epochs 1000 --init_lr 0.05
#####################################################
import sys, time, copy, torch, random, argparse
@ -41,7 +41,7 @@ def main(args):
w_container_per_epoch = dict()
per_timestamp_time, start_time = AverageMeter(), time.time()
for idx in range(1, env_info["total"]):
for idx in range(args.prev_time, env_info["total"]):
need_time = "Time Left: {:}".format(
convert_secs2time(per_timestamp_time.avg * (env_info["total"] - idx), True)
@ -53,8 +53,8 @@ def main(args):
+ need_time
)
# train the same data
historical_x = env_info["{:}-x".format(idx - 1)]
historical_y = env_info["{:}-y".format(idx - 1)]
historical_x = env_info["{:}-x".format(idx - args.prev_time)]
historical_y = env_info["{:}-y".format(idx - args.prev_time)]
# build model
model = get_model(**model_kwargs)
print(model)
@ -160,6 +160,12 @@ if __name__ == "__main__":
default=0.1,
help="The initial learning rate for the optimizer (default is Adam)",
)
parser.add_argument(
"--prev_time",
type=int,
default=5,
help="The gap between prev_time and current_timestamp",
)
parser.add_argument(
"--batch_size",
type=int,
@ -184,7 +190,12 @@ if __name__ == "__main__":
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
args.save_dir = "{:}-{:}-d{:}".format(
args.save_dir, args.env_version, args.hidden_dim
args.save_dir = "{:}-d{:}_e{:}_lr{:}-prev{:}-env{:}".format(
args.save_dir,
args.hidden_dim,
args.epochs,
args.init_lr,
args.prev_time,
args.env_version,
)
main(args)

6
exps/LFNA/basic-same.py
View File

@ -41,7 +41,7 @@ def main(args):
w_container_per_epoch = dict()
per_timestamp_time, start_time = AverageMeter(), time.time()
for idx in range(env_info["total"]):
for idx in range(1, env_info["total"]):
need_time = "Time Left: {:}".format(
convert_secs2time(per_timestamp_time.avg * (env_info["total"] - idx), True)
@ -184,7 +184,7 @@ if __name__ == "__main__":
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
args.save_dir = "{:}-{:}-d{:}".format(
args.save_dir, args.env_version, args.hidden_dim
args.save_dir = "{:}-d{:}_e{:}_lr{:}-env{:}".format(
args.save_dir, args.hidden_dim, args.epochs, args.init_lr, args.env_version
)
main(args)

12
exps/LFNA/lfna.py
View File

@ -157,11 +157,11 @@ def main(args):
per_epoch_time.update(time.time() - start_time)
start_time = time.time()
# meta-training
# meta-test
meta_model.load_best()
eval_env = env_info["dynamic_env"]
w_container_per_epoch = dict()
for idx in range(args.seq_length, env_info["total"]):
for idx in range(args.seq_length, len(eval_env)):
# build-timestamp
future_time = env_info["{:}-timestamp".format(idx)]
time_seqs = []
@ -176,8 +176,8 @@ def main(args):
future_container = seq_containers[-1]
w_container_per_epoch[idx] = future_container.no_grad_clone()
# evaluation
future_x = env_info["{:}-x".format(idx)]
future_y = env_info["{:}-y".format(idx)]
future_x = env_info["{:}-x".format(idx)].to(args.device)
future_y = env_info["{:}-y".format(idx)].to(args.device)
future_y_hat = base_model.forward_with_container(
future_x, w_container_per_epoch[idx]
)
@ -299,12 +299,12 @@ if __name__ == "__main__":
if args.rand_seed is None or args.rand_seed < 0:
args.rand_seed = random.randint(1, 100000)
assert args.save_dir is not None, "The save dir argument can not be None"
args.save_dir = "{:}-{:}-d{:}_{:}_{:}-e{:}".format(
args.save_dir = "{:}-d{:}_{:}_{:}-e{:}-env{:}".format(
args.save_dir,
args.env_version,
args.hidden_dim,
args.layer_dim,
args.time_dim,
args.epochs,
args.env_version,
)
main(args)

20
exps/LFNA/vis-synthetic.py
View File

@ -237,18 +237,20 @@ def compare_algs(save_dir, version, alg_dir="./outputs/lfna-synthetic"):
env_info = torch.load(cache_path)
alg_name2dir = OrderedDict()
alg_name2dir["Optimal"] = "use-same-timestamp"
# alg_name2dir["Supervised Learning (History Data)"] = "use-all-past-data"
# alg_name2dir["MAML"] = "use-maml-s1"
# alg_name2dir["LFNA (fix init)"] = "lfna-fix-init"
alg_name2dir["LFNA (debug)"] = "lfna-tall-hpnet"
alg_name2all_containers = OrderedDict()
if version == "v1":
poststr = "v1-d16"
# alg_name2dir["Optimal"] = "use-same-timestamp"
alg_name2dir["LFNA"] = "lfna-battle-v1-d16_16_16-e200"
alg_name2dir[
"Previous Timestamp"
] = "use-prev-timestamp-d16_e500_lr0.1-prev5-envv1"
else:
raise ValueError("Invalid version: {:}".format(version))
alg_name2all_containers = OrderedDict()
for idx_alg, (alg, xdir) in enumerate(alg_name2dir.items()):
ckp_path = Path(alg_dir) / "{:}-{:}".format(xdir, poststr) / "final-ckp.pth"
ckp_path = Path(alg_dir) / str(xdir) / "final-ckp.pth"
xdata = torch.load(ckp_path, map_location="cpu")
alg_name2all_containers[alg] = xdata["w_container_per_epoch"]
# load the basic model
@ -267,11 +269,11 @@ def compare_algs(save_dir, version, alg_dir="./outputs/lfna-synthetic"):
dynamic_env = env_info["dynamic_env"]
min_t, max_t = dynamic_env.min_timestamp, dynamic_env.max_timestamp
linewidths = 10
linewidths, skip = 10, 5
for idx, (timestamp, (ori_allx, ori_ally)) in enumerate(
tqdm(dynamic_env, ncols=50)
):
if idx == 0:
if idx <= skip:
continue
fig = plt.figure(figsize=figsize)
cur_ax = fig.add_subplot(2, 1, 1)
@ -335,9 +337,9 @@ def compare_algs(save_dir, version, alg_dir="./outputs/lfna-synthetic"):
cur_ax.set_ylim(0, 10)
cur_ax.legend(loc=1, fontsize=LegendFontsize)
pdf_save_path = save_dir / "pdf" / "v{:}-{:05d}.pdf".format(version, idx)
pdf_save_path = save_dir / "pdf" / "v{:}-{:05d}.pdf".format(version, idx - skip)
fig.savefig(str(pdf_save_path), dpi=dpi, bbox_inches="tight", format="pdf")
png_save_path = save_dir / "png" / "v{:}-{:05d}.png".format(version, idx)
png_save_path = save_dir / "png" / "v{:}-{:05d}.png".format(version, idx - skip)
fig.savefig(str(png_save_path), dpi=dpi, bbox_inches="tight", format="png")
plt.close("all")
save_dir = save_dir.resolve()

6
lib/datasets/synthetic_env.py
View File

@ -80,6 +80,12 @@ class SyntheticDEnv(data.Dataset):
def timestamp_interval(self):
return self._timestamp_generator.interval
def random_timestamp(self):
return (
random.random() * (self.max_timestamp - self.min_timestamp)
+ self.min_timestamp
)
def reset_max_seq_length(self, seq_length):
self._seq_length = seq_length

4
lib/datasets/synthetic_utils.py
View File

@ -56,11 +56,11 @@ class TimeStamp(UnifiedSplit, data.Dataset):
@property
def min_timestamp(self):
return self._min_timestamp
return self._min_timestamp + self._interval * min(self._indexes)
@property
def max_timestamp(self):
return self._max_timestamp
return self._min_timestamp + self._interval * max(self._indexes)
@property
def interval(self):