Merge pull request #535 from AIStream-Peelout/transformer_timeout2

Fix version errors final

.circleci/config.yml

@@ -6,7 +6,7 @@ orbs:
 executors:
   python-executor:
     docker:
-      - image: circleci/python:3.9.1
+      - image: cimg/python:3.10.3
     working_directory: ~/repo
 
 
@@ -35,8 +35,9 @@ jobs:
       - run:
           name: install dependencies
           command: |
-            sudo pip install flake8
-            flake8 .
+            python --version
+            pip install flake8
+            python -m flake8 .
 
   evaluator_test:
     <<: *defaults
@@ -50,10 +51,10 @@ jobs:
       - run:
           name: install dependencies
           command: |
-            sudo pip install pipenv
-            sudo pip install coverage
-            sudo pip install --quiet --force-reinstall  --upgrade -r requirements.txt
-            sudo python setup.py develop --no-deps
+             pip install pipenv
+             pip install coverage
+             pip install --quiet --force-reinstall  --upgrade -r requirements.txt
+             python setup.py develop --no-deps
 
       - save_cache:
           key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
@@ -93,10 +94,10 @@ jobs:
       - run:
           name: install dependencies
           command: |
-            sudo pip install pipenv
-            sudo pip install coverage
-            sudo pip install --force-reinstall --upgrade -r requirements.txt
-            sudo python setup.py develop --no-deps
+             pip install pipenv
+             pip install coverage
+             pip install --force-reinstall --upgrade -r requirements.txt
+             python setup.py develop --no-deps
 
       - save_cache:
           key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
@@ -139,10 +140,10 @@ jobs:
       - run:
           name: install dependencies
           command: |
-            sudo pip install pipenv
-            sudo pip install coverage
-            sudo pip install --force-reinstall --upgrade --quiet -r requirements.txt
-            sudo python setup.py develop --no-deps
+             pip install pipenv
+             pip install coverage
+             pip install --force-reinstall --upgrade --quiet -r requirements.txt
+             python setup.py develop --no-deps
 
       - save_cache:
           key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
@@ -163,8 +164,6 @@ jobs:
             bash <(curl -s https://codecov.io/bash) -cF python
             coverage run -m unittest -v tests/time_model_test.py
             bash <(curl -s https://codecov.io/bash) -cF python
-            coverage run -m unittest -v tests/model_config_tests.py
-            bash <(curl -s https://codecov.io/bash) -cF python
             coverage run -m unittest -v tests/data_loader_tests.py
             bash <(curl -s https://codecov.io/bash) -cF python
             coverage run -m unittest -v tests/test_da_rnn.py
@@ -200,10 +199,10 @@ jobs:
       - run:
           name: install dependencies
           command: |
-            sudo pip install pipenv
-            sudo pip install coverage
-            sudo pip install --upgrade --force-reinstall --quiet -r requirements.txt
-            sudo python setup.py develop --no-deps
+             pip install pipenv
+             pip install coverage
+             pip install --upgrade --force-reinstall --quiet -r requirements.txt
+             python setup.py develop --no-deps
 
       - save_cache:
           key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
@@ -239,10 +238,10 @@ jobs:
       - run:
           name: install dependencies
           command: |
-            sudo pip install pipenv
-            sudo pip install --force-reinstall --upgrade --quiet -r requirements.txt
-            sudo pip install coverage
-            sudo python setup.py develop --no-deps
+             pip install pipenv
+             pip install --force-reinstall --upgrade --quiet -r requirements.txt
+             pip install coverage
+             python setup.py develop --no-deps
       - save_cache:
           key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
           paths:
@@ -285,7 +284,44 @@ jobs:
             coverage run flood_forecast/trainer.py -p  tests/full_transformer.json
             echo -e 'running trainer_decoder_test \n'
             coverage run flood_forecast/trainer.py -p tests/decoder_test.json
-            echo -e 'running trainer_custom_encode \n'
+      - store_test_results:
+          path: test-results
+
+      - store_artifacts:
+          path: test-results
+          destination: test-results-trainer
+
+      - run:
+          name: upload results
+          when: always
+          command: bash <(curl -s https://codecov.io/bash) -cF python
+
+
+  trainer_test1:
+    <<: *defaults
+    steps:
+      - attach_workspace:
+          at: ~/repo
+      - restore_cache:  # ensure this step occurs *before* installing dependencies
+          key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
+
+      - run:
+          name: install dependencies
+          command: |
+             pip install pipenv
+             pip install --force-reinstall --upgrade --quiet -r requirements.txt
+             pip install coverage
+             python setup.py develop --no-deps
+      - save_cache:
+          key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
+          paths:
+            - ~/.local
+
+      - run:
+          name: Trainer1 tests
+          when: always
+          command: |
+          
             coverage run flood_forecast/trainer.py -p tests/custom_encode.json
             coverage run flood_forecast/trainer.py -p tests/multi_decoder_test.json
             coverage run flood_forecast/trainer.py -p tests/test_dual.json
@@ -295,7 +331,7 @@ jobs:
 
       - store_artifacts:
           path: test-results
-          destination: test-results-trainer
+          destination: test-results-trainer2
 
       - run:
           name: upload results
@@ -313,10 +349,10 @@ jobs:
       - run:
           name: install dependencies
           command: |
-            sudo pip install pipenv
-            sudo pip install coverage
-            sudo pip install --upgrade --force-reinstall --quiet -r requirements.txt
-            sudo python setup.py develop --no-deps
+             pip install pipenv
+             pip install coverage
+             pip install --upgrade --force-reinstall --quiet -r requirements.txt
+             python setup.py develop --no-deps
 
       - save_cache:
           key: requirements-v2-{{ .Branch }}-{{ checksum "requirements.txt" }}
@@ -370,3 +406,6 @@ workflows:
       - plot_test:
           requires:
             - linting
+      - trainer_test1:
+          requires:
+            - linting

.flake8

@@ -1,4 +1,4 @@
 [flake8]
-max_line_length=120
-ignore=E305,W504, E126
+max_line_length=121
+ignore=E305,W504,E126
 max-complexity=18

docs/source/buil_dataset.rst

@@ -1,4 +1,4 @@
-Build Dataset
+Build FF original Dataset
 =========================
 
 

flood_forecast/deployment/inference.py

@@ -68,13 +68,15 @@ def infer_now(self, some_date: datetime, csv_path=None, save_buck=None, save_nam
             self.inference_params["test_csv_path"] = csv_path
             self.inference_params["dataset_params"]["file_path"] = csv_path
         df, tensor, history, forecast_start, test, samples = infer_on_torch_model(self.model, **self.inference_params)
+        print("the tensor shape is 2 ")
+        print(tensor.shape)
         if test.scale and self.n_targets:
+            unscaled = test.inverse_scale(tensor.numpy())
             for i in range(0, self.n_targets):
-                unscaled = test.inverse_scale(tensor.numpy())
                 df["pred_" + self.targ_cols[i]] = 0
                 print("Shape of unscaled is: ")
                 print(unscaled.shape)
-                df["pred_" + self.targ_cols[i]][forecast_history:] = unscaled[0, :, i].numpy()
+                df["pred_" + self.targ_cols[i]][forecast_history:] = unscaled[:, i].numpy()
         elif test.scale:
             unscaled = test.inverse_scale(tensor.numpy().reshape(-1, 1))
             df["preds"][forecast_history:] = unscaled.numpy()[:, 0]

flood_forecast/preprocessing/pytorch_loaders.py

@@ -139,12 +139,18 @@ def inverse_scale(
             if len(result_data.shape) > 2:
                 result_data = result_data.permute(2, 0, 1).reshape(result_data.shape[2], -1)
                 result_data = result_data.permute(1, 0)
+            elif len(result_data.shape) > 1:
+                result_data = result_data
+            else:
+                result_data = result_data.unsqueeze(0)
             result_data_np = result_data.numpy()
         if isinstance(result_data, np.ndarray):
             result_data_np = result_data
         # print(type(result_data))
         if self.no_scale:
             return torch.from_numpy(result_data_np)
+        if len(result_data_np.shape) > 2:
+            result_data_np = result_data_np[0, :, :]
         return torch.from_numpy(
             self.targ_scaler.inverse_transform(result_data_np)
         )

flood_forecast/time_model.py

@@ -88,6 +88,7 @@ def upload_gcs(self, save_path: str, name: str, file_type: str, epoch=0, bucket_
     def wandb_init(self):
         if self.params["wandb"]:
             wandb.init(
+                id=wandb.util.generate_id(),
                 project=self.params["wandb"].get("project"),
                 config=self.params,
                 name=self.params["wandb"].get("name"),

requirements.txt

@@ -1,22 +1,23 @@
 shap==0.40.0
-scikit-learn==0.24.2
+scikit-learn>=1.0.1
 pandas
 torch
 tb-nightly
 seaborn
 future
 h5py
-wandb==0.12.15
+wandb==0.12.16
 google-cloud
 google-cloud-storage
 plotly~=5.6.0
-pytz~=2022.1
+pytz>=2022.1
 setuptools~=62.0.0
 numpy>=1.21
 requests
 torchvision>=0.6.0
-mpld3~=0.5
+mpld3>=0.5
 numba>=0.50
 sphinx
 sphinx-rtd-theme
 sphinx-autodoc-typehints
+sphinx

setup.py

@@ -16,7 +16,7 @@
 
 setup(
     name='flood_forecast',
-    version='0.995dev',
+    version='0.9956dev',
     packages=[
         'flood_forecast',
         'flood_forecast.transformer_xl',

tests/custom_encode.json

@@ -14,7 +14,7 @@
        "test_path": "tests/test_data/keag_small.csv",
        "forecast_history":5,
        "forecast_length":1,
-       "train_end": 100,
+       "train_end": 99,
        "valid_start":301,
        "valid_end": 401,
        "test_start":125,

tests/data_loader_tests.py

@@ -11,7 +11,7 @@
 
 class DataLoaderTests(unittest.TestCase):
     """
-    Class to test data loader functionality for the code.
+    Class to test data loader functionality for the code mod.
     Specifically, reuturn types and indexing to make sure there is no overlap.
     """
 

tests/decoder_test.json

@@ -12,7 +12,7 @@
        "training_path": "tests/test_data/keag_small.csv",
        "validation_path": "tests/test_data/keag_small.csv",
        "test_path": "tests/test_data/keag_small.csv",
-       "batch_size":4,
+       "batch_size":10,
        "forecast_history":5,
        "forecast_length":1,
        "train_end": 100,

tests/dsanet.json

@@ -20,7 +20,7 @@
       "training_path": "tests/test_data/keag_small.csv",
       "validation_path": "tests/test_data/keag_small.csv",
       "test_path": "tests/test_data/keag_small.csv",
-      "batch_size":4,
+      "batch_size":10,
       "forecast_history":10,
       "forecast_length":1,
       "train_start": 1,

tests/pytorc_train_tests.py

@@ -350,9 +350,11 @@ def test_handle_meta2(self):
     def test_scaling_data(self):
         scaled_src, _ = self.model.test_data[0]
         data_unscaled = self.model.test_data.original_df.iloc[0:20]["cfs"].values
+        print("shape bw")
+        print(scaled_src[:, 0].shape)
         inverse_scale = self.model.test_data.inverse_scale(scaled_src[:, 0])
-        self.assertAlmostEqual(inverse_scale.numpy()[0], data_unscaled[0])
-        self.assertAlmostEqual(inverse_scale.numpy()[9], data_unscaled[9])
+        self.assertAlmostEqual(inverse_scale.numpy()[0, 0], data_unscaled[0])
+        self.assertAlmostEqual(inverse_scale.numpy()[0, 9], data_unscaled[9])
 
     def test_compute_loss_no_scaling(self):
         exam = torch.Tensor([4.0]).repeat(2, 20, 5)

tests/test_explain_model_output.py

@@ -32,7 +32,7 @@ class ModelInterpretabilityTest(unittest.TestCase):
             "project": "repo-flood_forecast",
         },
         "inference_params": {
-            "hours_to_forecast": 30,
+            "hours_to_forecast": 32,
             "datetime_start": datetime(2014, 6, 2, 0),
         },
     }
@@ -54,7 +54,7 @@ class ModelInterpretabilityTest(unittest.TestCase):
             "project": "repo-flood_forecast",
         },
         "inference_params": {
-            "hours_to_forecast": 30,
+            "hours_to_forecast": 32,
             "datetime_start": datetime(2014, 6, 2, 0),
         },
     }

tests/test_inf_single.json

@@ -54,7 +54,7 @@
 
        },
        "lr": 0.4,
-       "epochs": 3,
+       "epochs": 1,
        "batch_size":4
 
     },

tests/test_informer.json

@@ -56,7 +56,7 @@
          "lr": 0.04
        },
 
-       "epochs": 6,
+       "epochs": 1,
        "batch_size":5
 
     },