Use poetry, make compatible newer lightning version

.gitignore

@@ -139,3 +139,5 @@ notebook/
 
 # Images
 *.jpg
+sandbox/
+lightning_logs/

detection/lightning/util.py

@@ -2,7 +2,8 @@
 import json
 import os
 from os.path import join
-from pytorch_lightning.loggers.test_tube import TestTubeLogger
+# from pytorch_lightning.loggers.test_tube import TestTubeLogger
+from pytorch_lightning.loggers import CSVLogger
 from pytorch_lightning.callbacks import ModelCheckpoint, EarlyStopping
 
 
@@ -12,12 +13,13 @@ def get_ckpt_dir(save_path, exp_name):
 
 def get_ckpt_callback(save_path, exp_name, monitor="val_loss", mode="min"):
     ckpt_dir = os.path.join(save_path, exp_name, "ckpts")
-    return ModelCheckpoint(filepath=ckpt_dir,
+    return ModelCheckpoint(dirpath=ckpt_dir,
                            save_top_k=1,
                            verbose=True,
                            monitor=monitor,
                            mode=mode,
-                           prefix='')
+                        #    prefix=''
+                           )
 
 
 def get_early_stop_callback(patience=10):
@@ -29,6 +31,6 @@ def get_early_stop_callback(patience=10):
 
 def get_logger(save_path, exp_name):
     exp_dir = os.path.join(save_path, exp_name)
-    return TestTubeLogger(save_dir=exp_dir,
+    return CSVLogger(save_dir=exp_dir,
                           name='lightning_logs',
                           version="0")

detection/main.py

@@ -20,7 +20,8 @@ def train(save_dir=C.SANDBOX_PATH,
           gpus=1,
           pretrained=True,
           batch_size=8,
-          accelerator="ddp",
+          accelerator="gpu",
+          strategy="ddp",
           gradient_clip_val=0.5,
           max_epochs=100,
           learning_rate=1e-5,
@@ -36,13 +37,14 @@ def train(save_dir=C.SANDBOX_PATH,
     Args:
         save_dir: Path to save the checkpoints and logs
         exp_name: Name of the experiment
-        model: Model name
+        model: Model name ("mask_rcnn","faster_rcnn","retinanet","rpn","fast_rcnn", see 'detection/models/detection/detectron.py')
         gpus: int. (ie: 2 gpus)
              OR list to specify which GPUs [0, 1] OR '0,1'
              OR '-1' / -1 to use all available gpus
         pretrained: Whether or not to use the pretrained model
         num_classes: Number of classes
-        accelerator: Distributed computing mode
+        accelerator: Supports passing different accelerator types (“cpu”, “gpu”, “tpu”, “ipu”, “hpu”, “mps, “auto”)
+        strategy:  Supports different training strategies with aliases as well custom strategies (e.g. "ddp")
         gradient_clip_val:  Clip value of gradient norm
         limit_train_batches: Proportion of training data to use
         max_epochs: Max number of epochs
@@ -63,15 +65,16 @@ def train(save_dir=C.SANDBOX_PATH,
     task = get_task(args)
     trainer = Trainer(gpus=gpus,
                       accelerator=accelerator,
+                      strategy=strategy,
                       logger=get_logger(save_dir, exp_name),
                       callbacks=[get_early_stop_callback(patience),
                                  get_ckpt_callback(save_dir, exp_name, monitor="mAP", mode="max")],
-                      weights_save_path=os.path.join(save_dir, exp_name),
+                      default_root_dir=os.path.join(save_dir, exp_name),
                       gradient_clip_val=gradient_clip_val,
                       limit_train_batches=limit_train_batches,
                       limit_val_batches=limit_val_batches,
                       limit_test_batches=limit_test_batches,
-                      weights_summary=weights_summary,
+                    #   weights_summary=weights_summary,
                       max_epochs=max_epochs)
     trainer.fit(task)
     return save_dir, exp_name

pyproject.toml

@@ -0,0 +1,52 @@
+[tool.poetry]
+name = "Surveilling-surveillance"
+version = "0.1.0"
+description = ""
+authors = ["Stanford Computational Policy Lab"]
+readme = "README.md"
+
+[tool.poetry.dependencies]
+# python can be ^3.7 if we describe the right toch dependencies
+#python = "~3.9"
+#this is basically the same, but != 3.9.7 because of dep 'streamlit'
+python = ">=3.9,<3.9.7 || >3.9.7,<3.10"
+# last version supporting np.object
+numpy = "1.23.4"
+#numpy = "^1.27.2"
+torch = [
+	{ version="1.12.1" },
+	{ url = "https://download.pytorch.org/whl/cu113/torch-1.12.1%2Bcu113-cp39-cp39-linux_x86_64.whl", markers = "python_version ~= '3.9' and sys_platform == 'linux'" }
+]
+torchvision = [
+	{ version="0.13.1" },
+	{ url = "https://download.pytorch.org/whl/cu113/torchvision-0.13.1%2Bcu113-cp39-cp39-linux_x86_64.whl", markers = "python_version ~= '3.9' and sys_platform == 'linux'" }
+]
+scikit-image = "^0.20.0"
+
+#PyYAML = "5.2"
+#scipy = "^1.10.0"
+#timm = "0.1.20"
+tensorboardx = "1.9"
+scikit-learn = "^1.3.0"
+#pytorch-lightning = "1.1.4"
+# work around https://lightning.ai/forums/t/attribute-error-no-sync-params/2258
+pytorch-lightning = "~1.8"
+test-tube = "0.7.1"
+tqdm = ">=4.36.1"
+pretrainedmodels = "^0.7.4"
+fire = "^0.5.0"
+streamlit = "^1.25.0"
+albumentations = "^1.3.1"
+imgaug = "^0.4.0"
+pytorch-ignite = "^0.4.12"
+seaborn = "^0.12.2"
+segmentation-models-pytorch = "^0.3.3"
+osmnx = "^1.6.0"
+geopy = "^2.3.0"
+coloredlogs = "^15.0.1"
+nni = "^3.0"
+pillow="~9.5"
+
+[build-system]
+requires = ["poetry-core"]
+build-backend = "poetry.core.masonry.api"