Updated a number of packages

.pre-commit-config.yaml

@@ -1,6 +1,6 @@
 repos:
 - repo: https://github.com/pycqa/isort
-  rev: 5.8.0
+  rev: 5.12.0
   hooks:
   - id: isort
     name: isort (python)

environment.yaml

@@ -4,10 +4,10 @@ channels:
   - conda-forge
 dependencies:
   - python=3.9
-  - gwpy=2.1.0
-  - astropy<5.0.0
-  - python-ldas-tools-framecpp
+  - gwpy=2.1.5
+  - astropy
+  - python-ldas-tools-framecpp<3.0.0
   - python-nds2-client
-  - pytorch
+  - pytorch=1.13.1
   - torchvision
   - cudatoolkit=11.3

libs/trainer/pyproject.toml

@@ -10,7 +10,7 @@ packages = [
 
 [tool.poetry.dependencies]
 python = "^3.8,<3.10"
-torch = "^1.0"
+torch = {version = "^1.13", source = "torch"}
 
 # include signal processing libraries
 # for preprocessing steps
@@ -26,6 +26,12 @@ torch = "^1.0"
 
 "hermes.typeo" = "^0.1.5"
 
+[[tool.poetry.source]]
+name = "torch"
+url = "https://download.pytorch.org/whl/cu116"
+secondary = true
+default = false
+
 [tool.poetry.dev-dependencies]
 pytest = "^6.2"
 

projects/sandbox/README.md

@@ -17,10 +17,10 @@ Compute and plot the ASDs of the raw and cleaned frames, as well as the ratio be
 ## Running
 If you installed the `pinto` command line utility described in the root [README](../../README.md), you can build each component's environment individually by running  `pinto build <path to component>`. Consult each component's documentation for the commands and options they expose.
 
-Running the workflow end-to-end requires exporting three environment variables (with apologies for the inconsistent naming conventions for now)
+Running the workflow end-to-end requires exporting three environment variables
 
 ```console
-PROJECT_DIRECTORY="..."  # the directory to write workflow outputs to
+PROJECT_DIR="..."  # the directory to write workflow outputs to
 DATA_DIR="..."  # the path to a directory of 1s .gwf witness/strain files
 MODEL_REPOSITORY="..."  # the directory to serve models from
 ```

projects/sandbox/export/export.py

@@ -227,7 +227,8 @@ def export(
     logging.info(f"Creating model and loading weights from {weights}")
     nn = architecture(len(channels) - 1)
     nn = PrePostDeepClean(nn)
-    nn.load_state_dict(torch.load(weights))
+    state_dict = torch.load(weights, map_location="cpu")
+    nn.load_state_dict(state_dict)
     nn.eval()
 
     # instantiate a model repository at the