clean group weight

configs/unifuse_st3d.cfg

@@ -1,5 +1,6 @@
 [EXP_SETTING]
-id=unifuse_st3d_pilot
+# id=unifuse_st3d_pilot
+id=unifuse_st3d_pilot_group
 
 # #### Dataset & Augmentation setting ####
 root=/project/albert/datasets/planar/mp3d
@@ -15,8 +16,6 @@ unlabel_train_txt=/project/albert/datasets/planar/st3d/st3d_all.txt
 zero_shot_root=/project/albert/datasets/planar/stanford2D3D
 zero_shot_txt=/project/albert/datasets/planar/sf3d_txt/test.txt
 
-# relative or metric
-# relative=CUBE
 median_align=True
 dataset=mp3d
 epochs=200

train.py

@@ -320,15 +320,14 @@ def val(
     epoch: int, 
     writer: SummaryWriter=None, 
     evaluator: Affine_Inv_Evaluator=None, 
-    mode='Valid', 
-    save_log_flag: bool=True):
+    mode='Valid'):
     """Eval function."""
     model.eval()
     with torch.no_grad():
         pbar = tqdm(dataloader)
         pbar.set_description("{} Epoch_{}".format(mode, epoch))
         total_loss = defaultdict(float)
-        for batch_idx, inputs in enumerate(pbar):
+        for _, inputs in enumerate(pbar):
             for key, ipt in inputs.items():
                 if key not in ["rgb", "rgb_name"]:
                     inputs[key] = ipt.to(args.device)
@@ -352,11 +351,10 @@ def val(
                 total_loss[f'total_{k}'] += v.data.cpu().numpy() / len(dataloader)
 
         # Evaluator to shell and tensorboard
-        for i, key in enumerate(evaluator.metrics.keys()):
+        for _, key in enumerate(evaluator.metrics.keys()):
             total_loss[f'total_{key}'] = np.array(evaluator.metrics[key].avg.cpu())
         evaluator.print()
-        if save_log_flag:
-            save_log(writer, inputs, outputs, total_loss, args)
+        save_log(writer, inputs, outputs, total_loss, args)
 
 
 def main_joint_unlabel():
@@ -410,7 +408,7 @@ def main_joint_unlabel():
             val(args, model, test_loader, epoch, writer['test'], evaluator, mode='Test')
         if zeroshot_loader is not None:
             zeroshot_evaluator.reset_eval_metrics()
-            val(args, model, zeroshot_loader, epoch, writer['zeroshot'], zeroshot_evaluator, mode='zeroshot')
+            val(args, model, zeroshot_loader, epoch, writer['zeroshot'], zeroshot_evaluator, mode='Zeroshot')
 
         if (epoch + 1) % args.save_every == 0:
             save_model(model, optimizer, args)

utils/get_args.py

@@ -1,9 +1,5 @@
-import os
-import sys
-import time
 import argparse
 import configparser
-from threading import Thread
 
 
 def force_config_value_type(val):
@@ -102,7 +98,6 @@ def parse_args():
     """
     Pre-processing setting
     """
-    # parser.add_argument('--base_height', default=512, type=int)
     parser.add_argument('--h', default=512, type=int, help='loader process height')
     parser.add_argument('--w', default=1024, type=int, help='loader process width')
     parser.add_argument('--rgb_mean', default=[0.485, 0.456, 0.406], nargs=3, type=float)

utils/utils.py

@@ -1,8 +1,3 @@
-import os
-import pickle
-
-import torch
-import torch.nn as nn
 import numpy as np
 
 
@@ -16,13 +11,11 @@ def read_list(list_file):
 
 
 def read_list_with_ndarray(list_file):
-    # rgb_depth_list = np.empty((0, 2))
     rgb_list = np.empty(0)
     depth_list = np.empty(0)
     with open(list_file) as f:
         lines = f.readlines()
         for line in lines:
-            # rgb_depth_list = np.append(rgb_depth_list, [line.strip().split(" ")], 0)
             rgb, depth = line.strip().split(" ")
             rgb_list = np.append(rgb_list, rgb)
             depth_list = np.append(depth_list, depth)
@@ -36,85 +29,13 @@ def read_list_with_ndarray(list_file):
 def group_weight(module):
     group_decay = []
     group_no_decay = []
-    # for m in module.modules():
-    #     if isinstance(m, nn.Linear):
-    #         group_decay.append(m.weight)
-    #         if m.bias is not None:
-    #             group_no_decay.append(m.bias)
-    #     elif isinstance(m, nn.modules.conv._ConvNd):
-    #         group_decay.append(m.weight)
-    #         if m.bias is not None:
-    #             group_no_decay.append(m.bias)
-    #     elif isinstance(m, nn.modules.batchnorm._BatchNorm):
-    #         if m.weight is not None:
-    #             group_no_decay.append(m.weight)
-    #         if m.bias is not None:
-    #             group_no_decay.append(m.bias)
-    #     elif isinstance(m, nn.GroupNorm):
-    #         if m.weight is not None:
-    #             group_no_decay.append(m.weight)
-    #         if m.bias is not None:
-    #             group_no_decay.append(m.bias)
-    #     elif isinstance(m, nn.LayerNorm):
-    #         if m.weight is not None:
-    #             group_no_decay.append(m.weight)
-    #         if m.bias is not None:
-    #             group_no_decay.append(m.bias)
-    #     else:
-    #         import pdb
-    #         # if m.bias is not None:
-    #         #     group_no_decay.append(m.bias)
 
     for name, param in module.named_parameters():
         if name.endswith('weight'):
             group_decay.append(param)
         elif name.endswith('bias'):
             group_no_decay.append(param)
-    # assert len(list(module.parameters())) == len(group_decay) + len(group_no_decay)
     return [
         dict(params=[param for param in group_decay if param.requires_grad]),
         dict(params=[param for param in group_no_decay if param.requires_grad], weight_decay=.0)
     ]
-
-
-# Some recommendations for memory leak (not solution for this time)
-# https://github.com/Lightning-AI/pytorch-lightning/issues/17257
-# https://github.com/ppwwyyxx/RAM-multiprocess-dataloader/blob/79897b26a2c4185a3ed086f18be5ea300913d5b7/serialize.py#L40-L50
-class NumpySerializedList():
-    def __init__(self, lst: list):
-        def _serialize(data):
-            buffer = pickle.dumps(data, protocol=-1)
-            return np.frombuffer(buffer, dtype=np.uint8)
-
-        print(
-            "Serializing {} elements to byte tensors and concatenating them all ...".format(
-                len(lst)
-            )
-        )
-        self._lst = [_serialize(x) for x in lst]
-        self._addr = np.asarray([len(x) for x in self._lst], dtype=np.int64)
-        self._addr = np.cumsum(self._addr)
-        self._lst = np.concatenate(self._lst)
-        print("Serialized dataset takes {:.2f} MiB".format(len(self._lst) / 1024**2))
-
-    def __len__(self):
-        return len(self._addr)
-
-    def __getitem__(self, idx):
-        start_addr = 0 if idx == 0 else self._addr[idx - 1].item()
-        end_addr = self._addr[idx].item()
-        bytes = memoryview(self._lst[start_addr:end_addr])
-        return pickle.loads(bytes)
-
-
-class TorchSerializedList(NumpySerializedList):
-    def __init__(self, lst: list):
-        super().__init__(lst)
-        self._addr = torch.from_numpy(self._addr)
-        self._lst = torch.from_numpy(self._lst)
-
-    def __getitem__(self, idx):
-        start_addr = 0 if idx == 0 else self._addr[idx - 1].item()
-        end_addr = self._addr[idx].item()
-        bytes = memoryview(self._lst[start_addr:end_addr].numpy())
-        return pickle.loads(bytes)