Fix bugs in LFNA

.latent-data/init-configs/setup.sh

@@ -7,6 +7,7 @@ echo "script-directory: $script_dir"
 cp ${script_dir}/tmux.conf ~/.tmux.conf
 cp ${script_dir}/vimrc ~/.vimrc
 cp ${script_dir}/bashrc ~/.bashrc
+cp ${script_dir}/condarc ~/.condarc
 
 wget https://repo.anaconda.com/miniconda/Miniconda3-4.7.12.1-Linux-x86_64.sh
 wget https://repo.anaconda.com/archive/Anaconda3-2020.11-Linux-x86_64.sh

exps/LFNA/basic-his.py

@@ -161,10 +161,7 @@ if __name__ == "__main__":
         help="The synthetic enviornment version.",
     )
     parser.add_argument(
-        "--hidden_dim",
+        "--hidden_dim", type=int, required=True, help="The hidden dimension.",
-        type=int,
-        required=True,
-        help="The hidden dimension.",
     )
     parser.add_argument(
         "--init_lr",
@@ -173,16 +170,10 @@ if __name__ == "__main__":
         help="The initial learning rate for the optimizer (default is Adam)",
     )
     parser.add_argument(
-        "--batch_size",
+        "--batch_size", type=int, default=512, help="The batch size",
-        type=int,
-        default=512,
-        help="The batch size",
     )
     parser.add_argument(
-        "--epochs",
+        "--epochs", type=int, default=1000, help="The total number of epochs.",
-        type=int,
-        default=1000,
-        help="The total number of epochs.",
     )
     parser.add_argument(
         "--srange", type=str, required=True, help="The range of models to be evaluated"

exps/LFNA/basic-maml.py

@@ -41,10 +41,7 @@ class MAML:
         )
         self.meta_lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(
             self.meta_optimizer,
-            milestones=[
+            milestones=[int(epochs * 0.8), int(epochs * 0.9),],
-                int(epochs * 0.8),
-                int(epochs * 0.9),
-            ],
             gamma=0.1,
         )
         self.inner_lr = inner_lr
@@ -197,10 +194,7 @@ if __name__ == "__main__":
         help="The synthetic enviornment version.",
     )
     parser.add_argument(
-        "--hidden_dim",
+        "--hidden_dim", type=int, default=16, help="The hidden dimension.",
-        type=int,
-        default=16,
-        help="The hidden dimension.",
     )
     parser.add_argument(
         "--meta_lr",
@@ -230,16 +224,10 @@ if __name__ == "__main__":
         help="The gap between prev_time and current_timestamp",
     )
     parser.add_argument(
-        "--meta_batch",
+        "--meta_batch", type=int, default=64, help="The batch size for the meta-model",
-        type=int,
-        default=64,
-        help="The batch size for the meta-model",
     )
     parser.add_argument(
-        "--epochs",
+        "--epochs", type=int, default=2000, help="The total number of epochs.",
-        type=int,
-        default=2000,
-        help="The total number of epochs.",
     )
     parser.add_argument(
         "--early_stop_thresh",

exps/LFNA/basic-prev.py

@@ -149,10 +149,7 @@ if __name__ == "__main__":
         help="The synthetic enviornment version.",
     )
     parser.add_argument(
-        "--hidden_dim",
+        "--hidden_dim", type=int, required=True, help="The hidden dimension.",
-        type=int,
-        required=True,
-        help="The hidden dimension.",
     )
     parser.add_argument(
         "--init_lr",
@@ -167,16 +164,10 @@ if __name__ == "__main__":
         help="The gap between prev_time and current_timestamp",
     )
     parser.add_argument(
-        "--batch_size",
+        "--batch_size", type=int, default=512, help="The batch size",
-        type=int,
-        default=512,
-        help="The batch size",
     )
     parser.add_argument(
-        "--epochs",
+        "--epochs", type=int, default=300, help="The total number of epochs.",
-        type=int,
-        default=300,
-        help="The total number of epochs.",
     )
     parser.add_argument(
         "--workers",

exps/LFNA/basic-same.py

@@ -149,10 +149,7 @@ if __name__ == "__main__":
         help="The synthetic enviornment version.",
     )
     parser.add_argument(
-        "--hidden_dim",
+        "--hidden_dim", type=int, required=True, help="The hidden dimension.",
-        type=int,
-        required=True,
-        help="The hidden dimension.",
     )
     parser.add_argument(
         "--init_lr",
@@ -161,16 +158,10 @@ if __name__ == "__main__":
         help="The initial learning rate for the optimizer (default is Adam)",
     )
     parser.add_argument(
-        "--batch_size",
+        "--batch_size", type=int, default=512, help="The batch size",
-        type=int,
-        default=512,
-        help="The batch size",
     )
     parser.add_argument(
-        "--epochs",
+        "--epochs", type=int, default=300, help="The total number of epochs.",
-        type=int,
-        default=300,
-        help="The total number of epochs.",
     )
     parser.add_argument(
         "--workers",

exps/LFNA/lfna-debug-hpnet.py

@@ -62,10 +62,7 @@ def main(args):
     )
     lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(
         optimizer,
-        milestones=[
+        milestones=[int(args.epochs * 0.8), int(args.epochs * 0.9),],
-            int(args.epochs * 0.8),
-            int(args.epochs * 0.9),
-        ],
         gamma=0.1,
     )
 
@@ -173,10 +170,7 @@ if __name__ == "__main__":
         help="The synthetic enviornment version.",
     )
     parser.add_argument(
-        "--hidden_dim",
+        "--hidden_dim", type=int, required=True, help="The hidden dimension.",
-        type=int,
-        required=True,
-        help="The hidden dimension.",
     )
     #####
     parser.add_argument(
@@ -186,10 +180,7 @@ if __name__ == "__main__":
         help="The initial learning rate for the optimizer (default is Adam)",
     )
     parser.add_argument(
-        "--meta_batch",
+        "--meta_batch", type=int, default=64, help="The batch size for the meta-model",
-        type=int,
-        default=64,
-        help="The batch size for the meta-model",
     )
     parser.add_argument(
         "--early_stop_thresh",
@@ -198,22 +189,13 @@ if __name__ == "__main__":
         help="The maximum epochs for early stop.",
     )
     parser.add_argument(
-        "--epochs",
+        "--epochs", type=int, default=2000, help="The total number of epochs.",
-        type=int,
-        default=2000,
-        help="The total number of epochs.",
     )
     parser.add_argument(
-        "--per_epoch_step",
+        "--per_epoch_step", type=int, default=20, help="The total number of epochs.",
-        type=int,
-        default=20,
-        help="The total number of epochs.",
     )
     parser.add_argument(
-        "--device",
+        "--device", type=str, default="cpu", help="",
-        type=str,
-        default="cpu",
-        help="",
     )
     # Random Seed
     parser.add_argument("--rand_seed", type=int, default=-1, help="manual seed")

exps/LFNA/lfna.py

@@ -101,10 +101,7 @@ def main(args):
     )
     lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(
         optimizer,
-        milestones=[
+        milestones=[int(args.epochs * 0.8), int(args.epochs * 0.9),],
-            int(args.epochs * 0.8),
-            int(args.epochs * 0.9),
-        ],
         gamma=0.1,
     )
     logger.log("The base-model is\n{:}".format(base_model))
@@ -166,7 +163,7 @@ def main(args):
     w_container_per_epoch = dict()
     for idx in range(args.seq_length, len(eval_env)):
         # build-timestamp
-        future_time = env_info["{:}-timestamp".format(idx)]
+        future_time = env_info["{:}-timestamp".format(idx)].item()
         time_seqs = []
         for iseq in range(args.seq_length):
             time_seqs.append(future_time - iseq * eval_env.timestamp_interval)
@@ -190,7 +187,7 @@ def main(args):
             )
 
         # creating the new meta-time-embedding
-        distance = meta_model.get_closest_meta_distance(future_time.item())
+        distance = meta_model.get_closest_meta_distance(future_time)
         if distance < eval_env.timestamp_interval:
             continue
         #
@@ -198,7 +195,9 @@ def main(args):
         optimizer = torch.optim.Adam(
             [new_param], lr=args.init_lr, weight_decay=1e-5, amsgrad=True
         )
-        meta_model.replace_append_learnt(torch.Tensor([future_time]).to(args.device), new_param)
+        meta_model.replace_append_learnt(
+            torch.Tensor([future_time], device=args.device), new_param
+        )
         meta_model.eval()
         base_model.train()
         for iepoch in range(args.epochs):
@@ -241,22 +240,13 @@ if __name__ == "__main__":
         help="The synthetic enviornment version.",
     )
     parser.add_argument(
-        "--hidden_dim",
+        "--hidden_dim", type=int, default=16, help="The hidden dimension.",
-        type=int,
-        default=16,
-        help="The hidden dimension.",
     )
     parser.add_argument(
-        "--layer_dim",
+        "--layer_dim", type=int, default=16, help="The layer chunk dimension.",
-        type=int,
-        default=16,
-        help="The layer chunk dimension.",
     )
     parser.add_argument(
-        "--time_dim",
+        "--time_dim", type=int, default=16, help="The timestamp dimension.",
-        type=int,
-        default=16,
-        help="The timestamp dimension.",
     )
     #####
     parser.add_argument(
@@ -272,10 +262,7 @@ if __name__ == "__main__":
         help="The weight decay for the optimizer (default is Adam)",
     )
     parser.add_argument(
-        "--meta_batch",
+        "--meta_batch", type=int, default=64, help="The batch size for the meta-model",
-        type=int,
-        default=64,
-        help="The batch size for the meta-model",
     )
     parser.add_argument(
         "--sampler_enlarge",
@@ -297,10 +284,7 @@ if __name__ == "__main__":
         "--workers", type=int, default=4, help="The number of workers in parallel."
     )
     parser.add_argument(
-        "--device",
+        "--device", type=str, default="cpu", help="",
-        type=str,
-        default="cpu",
-        help="",
     )
     # Random Seed
     parser.add_argument("--rand_seed", type=int, default=-1, help="manual seed")

exps/LFNA/lfna_meta_model.py

@@ -75,8 +75,7 @@ class LFNA_Meta(super_core.SuperModule):
 
         # unknown token
         self.register_parameter(
-            "_unknown_token",
+            "_unknown_token", torch.nn.Parameter(torch.Tensor(1, time_embedding)),
-            torch.nn.Parameter(torch.Tensor(1, time_embedding)),
         )
 
         # initialization

exps/LFNA/vis-synthetic.py

@@ -164,10 +164,8 @@ def compare_cl(save_dir):
         )
     print("Save all figures into {:}".format(save_dir))
     save_dir = save_dir.resolve()
-    base_cmd = (
+    base_cmd = "ffmpeg -y -i {xdir}/%04d.png -vf fps=1 -vf scale=2200:1800 -vb 5000k".format(
-        "ffmpeg -y -i {xdir}/%04d.png -vf fps=1 -vf scale=2200:1800 -vb 5000k".format(
+        xdir=save_dir
-            xdir=save_dir
-        )
     )
     video_cmd = "{:} -pix_fmt yuv420p {xdir}/compare-cl.mp4".format(
         base_cmd, xdir=save_dir