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

Upgrade lfna debug

Browse Source
This commit is contained in:
D-X-Y
2021-05-12 09:36:04 +00:00
parent f1c47af5fa
commit 4c51f62906
2 changed files with 36 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
exps/LFNA/lfna-tall-hpnet.py
View File

@@ -1,7 +1,7 @@
#####################################################
# Copyright (c) Xuanyi Dong [GitHub D-X-Y], 2021.04 #
#####################################################
# python exps/LFNA/lfna-tall-hpnet.py --env_version v1 --hidden_dim 16
# python exps/LFNA/lfna-tall-hpnet.py --env_version v1 --hidden_dim 16 --epochs 100000 --meta_batch 16
#####################################################
import sys, time, copy, torch, random, argparse
from tqdm import tqdm
@@ -42,7 +42,7 @@ def main(args):
hypernet = HyperNet(shape_container, args.hidden_dim, args.task_dim)
total_bar = env_info["total"] - 1
task_embeds = []
for i in range(total_bar):
for i in range(env_info["total"]):
task_embeds.append(torch.nn.Parameter(torch.Tensor(1, args.task_dim)))
for task_embed in task_embeds:
trunc_normal_(task_embed, std=0.02)
@@ -97,7 +97,7 @@ def main(args):
if iepoch % 200 == 0:
logger.log(
head_str
+ "meta-loss: {:.4f} ({:.4f}) :: lr={:.5f}, batch={:}, limit={:}".format(
+ " meta-loss: {:.4f} ({:.4f}) :: lr={:.5f}, batch={:}, limit={:}".format(
loss_meter.avg,
loss_meter.val,
min(lr_scheduler.get_last_lr()),
@@ -109,7 +109,7 @@ def main(args):
save_checkpoint(
{
"hypernet": hypernet.state_dict(),
"task_embed": task_embed,
"task_embeds": task_embeds,
"lr_scheduler": lr_scheduler.state_dict(),
"iepoch": iepoch,
},
@@ -122,6 +122,25 @@ def main(args):
print(model)
print(hypernet)
w_container_per_epoch = dict()
for idx in range(0, env_info["total"]):
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])
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()

17
exps/LFNA/lfna-test-hpnet.py
View File

@@ -34,17 +34,20 @@ 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)
model = model.to(args.device)
criterion = torch.nn.MSELoss()
logger.log("There are {:} weights.".format(model.get_w_container().numel()))
shape_container = model.get_w_container().to_shape_container()
hypernet = HyperNet(shape_container, args.hidden_dim, args.task_dim)
hypernet = hypernet.to(args.device)
# task_embed = torch.nn.Parameter(torch.Tensor(env_info["total"], args.task_dim))
total_bar = 10
task_embeds = []
for i in range(total_bar):
task_embeds.append(torch.nn.Parameter(torch.Tensor(1, args.task_dim)))
tensor = torch.Tensor(1, args.task_dim).to(args.device)
task_embeds.append(torch.nn.Parameter(tensor))
for task_embed in task_embeds:
trunc_normal_(task_embed, std=0.02)
@@ -79,8 +82,8 @@ def main(args):
# cur_time = random.randint(0, total_bar)
cur_task_embed = task_embeds[cur_time]
cur_container = hypernet(cur_task_embed)
cur_x = env_info["{:}-x".format(cur_time)]
cur_y = env_info["{:}-y".format(cur_time)]
cur_x = env_info["{:}-x".format(cur_time)].to(args.device)
cur_y = env_info["{:}-y".format(cur_time)].to(args.device)
cur_dataset = TimeData(cur_time, cur_x, cur_y)
preds = model.forward_with_container(cur_dataset.x, cur_container)
@@ -98,7 +101,7 @@ def main(args):
if iepoch % 200 == 0:
logger.log(
head_str
+ "meta-loss: {:.4f} ({:.4f}) :: lr={:.5f}, batch={:}".format(
+ " meta-loss: {:.4f} ({:.4f}) :: lr={:.5f}, batch={:}".format(
loss_meter.avg,
loss_meter.val,
min(lr_scheduler.get_last_lr()),
@@ -166,6 +169,12 @@ if __name__ == "__main__":
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()