v1.1.4: Retries, Fix Instance Destroy, Upgraded requirements.txt, and Latest Python

.github/workflows/test-and-deploy.yml

@@ -7,33 +7,24 @@ on:
 concurrency: deploy-numerai-cli-${{ github.ref }}
 
 jobs:
-  build-python36:
-    name: Build Python 3.6
-    runs-on: ubuntu-20.04
+  build-python:
+    name: Build Python ${{ matrix.python-version }}
+    strategy:
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
+    runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.6.15"
-      - run: pip install .
-      - run: numerai copy-example
-      - run: test -e tournament-python3/predict.py
-
-  build-python37:
-    name: Build Python 3.7
-    runs-on: ubuntu-20.04
-    steps:
-      - uses: actions/checkout@v2
-      - uses: actions/setup-python@v4
-        with:
-          python-version: "3.7.9"
+          python-version: ${{ matrix.python-version }}
       - run: pip install .
       - run: numerai copy-example
       - run: test -e tournament-python3/predict.py
 
   build-docker:
     name: Build Docker
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
       - run: |
@@ -44,14 +35,14 @@ jobs:
 
   pypi-release:
     name: PyPI Release
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     if: github.ref == 'refs/heads/master'
-    needs: [build-python36, build-python37, build-docker]
+    needs: [build-python, build-docker]
     steps:
       - uses: actions/checkout@v2
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.7.9"
+          python-version: "3.12"
       - run: pip install twine wheel
       - run: python setup.py sdist bdist_wheel
       - run: |

.gitignore

@@ -114,3 +114,7 @@ terraform.tfstate*
 
 .idea/
 .DS_STORE
+
+tournament-python*/
+signals-python*/
+crypto-python*/

numerai/cli/destroy_all.py

@@ -62,7 +62,7 @@ def destroy_all(ctx, verbose, preserve_node_config):
                 verbose,
                 provider,
                 env_vars=provider_keys,
-                inputs={"node_config_file": "nodes.json"},
+                inputs={"node_config_file": "../nodes.json"},
             )
 
     except Exception as e:

numerai/cli/node/destroy.py

@@ -62,7 +62,7 @@ def destroy(ctx, preserve_node_config, verbose):
             verbose,
             provider,
             env_vars=provider_keys,
-            inputs={"node_config_file": "nodes.json"},
+            inputs={"node_config_file": "../nodes.json"},
         )
 
     except Exception as e:

numerai/cli/uninstall.py

@@ -55,7 +55,7 @@ def uninstall():
                         provider=provider,
                         verbose=True,
                         env_vars=provider_keys,
-                        inputs={"node_config_file": "nodes.json"},
+                        inputs={"node_config_file": "../nodes.json"},
                     )
 
             click.secho("cleaning up docker images...")

numerai/cli/upgrade.py

@@ -210,7 +210,7 @@ def upgrade(verbose):
             verbose,
             provider="aws",
             env_vars=load_or_init_keys("aws"),
-            inputs={"node_config_file": "nodes.json"},
+            inputs={"node_config_file": "../nodes.json"},
         )
 
     click.secho("Upgrade complete!", fg="green")

numerai/cli/util/debug.py

@@ -151,6 +151,18 @@ def root_cause(std_out, err_msg):
             "mode: https://docs.docker.com/engine/security/rootless/"
         )
 
+    if (
+        "RepositoryAlreadyExistsException" in all_logs
+        or "EntityAlreadyExists" in all_logs
+        or "ResourceAlreadyExistsException" in all_logs
+    ):
+        raise exception_with_msg(
+            "AWS resources with the names this node expects already exist in your account, "
+            "but they are not tracked in the current local Terraform state. "
+            "This usually happens after a partial previous apply. "
+            "Delete or import the existing AWS resources shown above, then retry."
+        )
+
     # these are non-errors that either shouldn't be handled or are handled elsewhere
     if b"Can't update submission after deadline" in err_msg:
         return

numerai/cli/util/docker.py

@@ -69,7 +69,6 @@ def execute(command, verbose, censor_substr=None):
         shell=True,
         stdout=subprocess.PIPE,
         stderr=subprocess.PIPE,
-        bufsize=1,
         close_fds=on_posix,
     )
     stdout_q = Queue()

numerai/cli/util/terraform.py

@@ -19,7 +19,7 @@ def apply_terraform(nodes_config, affected_providers, provider, verbose):
                 verbose,
                 affected_provider,
                 env_vars=load_or_init_keys(affected_provider),
-                inputs={"node_config_file": "nodes.json"},
+                inputs={"node_config_file": "../nodes.json"},
             )
         else:
             click.secho(f"provider {affected_provider} not supported", fg="red")
@@ -51,7 +51,7 @@ def create_azure_registry(provider, provider_keys, verbose):
         verbose,
         "azure",
         env_vars=provider_keys,
-        inputs={"node_config_file": "nodes.json"},
+        inputs={"node_config_file": "../nodes.json"},
     )
     res = terraform("output -json acr_repo_details", True, provider).decode("utf-8")
     return json.loads(res)
@@ -64,13 +64,13 @@ def create_gcp_registry(provider, verbose):
         'apply -target="google_project_service.cloud_resource_manager" -auto-approve ',
         verbose,
         "gcp",
-        inputs={"node_config_file": "nodes.json"},
+        inputs={"node_config_file": "../nodes.json"},
     )
     terraform(
         'apply -target="google_artifact_registry_repository.registry[0]" -auto-approve ',
         verbose,
         "gcp",
-        inputs={"node_config_file": "nodes.json"},
+        inputs={"node_config_file": "../nodes.json"},
     )
     res = terraform("output -json artifact_registry_details", True, provider).decode(
         "utf-8"

numerai/examples/crypto-python3/Dockerfile

@@ -1,5 +1,5 @@
 # Provides us a working Python 3 environment.
-FROM python:3.9
+FROM python:3.13
 
 # These are docker arguments that `numerai node deploy/test` will always pass into docker.
 # They are then set in your environment so that numerapi can access them when uploading submissions.

numerai/examples/crypto-python3/predict.py

@@ -11,7 +11,8 @@
 logging.basicConfig(filename="log.txt", filemode="a")
 
 TOURNAMENT = 12
-DATA_VERSION = "crypto/v1.0"
+DATA_VERSION = "crypto/v2.0"
+TARGET_COL = "target_binned_return_20"
 TRAINED_MODEL_PREFIX = "./trained_model"
 
 DEFAULT_MODEL_ID = None
@@ -22,7 +23,7 @@
 # these are set by the docker image that you deploy after training,
 # but you can also set them manually above for local testing
 MODEL_ID = os.getenv("MODEL_ID", DEFAULT_MODEL_ID)
-napi = numerapi.NumerAPI(
+napi = numerapi.CryptoAPI(
     public_id=os.getenv("NUMERAI_PUBLIC_ID", DEFAULT_PUBLIC_ID),
     secret_key=os.getenv("NUMERAI_SECRET_KEY", DEFAULT_SECRET_KEY),
 )
@@ -40,63 +41,44 @@ def train(napi, model_id, force_training=False):
         return model
 
     logging.info("reading training data")
-    napi.download_dataset(f"{DATA_VERSION}/train_targets.parquet")
-    target = pd.read_parquet(f"{DATA_VERSION}/train_targets.parquet")
-
-    # TODO: implement get_features and train a model
-    # This will take a few minutes 🍵
-    # logging.info("training model")
-    # model = lgbm.LGBMRegressor(
-    #     n_estimators=2000,
-    #     learning_rate=0.01,
-    #     max_depth=5,
-    #     num_leaves=2**5-1,
-    #     colsample_bytree=0.1
-    # )
-    # model.fit(
-    #     train_data[feature_cols],
-    #     train_data["target"]
-    # )
-
-    # logging.info("saving model")
-    # joblib.dump(model, model_name)
-    # return model
-
-    # just return the target for now
-    return target
+    napi.download_dataset(f"{DATA_VERSION}/train.parquet")
+    train_data = pd.read_parquet(f"{DATA_VERSION}/train.parquet")
+    feature_cols = [col for col in train_data.columns if col.startswith("feature_")]
+
+    logging.info("training model")
+    model = lgbm.LGBMRegressor(
+        n_estimators=2000,
+        learning_rate=0.01,
+        max_depth=5,
+        num_leaves=2**5 - 1,
+        colsample_bytree=0.1,
+    )
+    model.fit(train_data[feature_cols], train_data[TARGET_COL])
+
+    logging.info("saving model")
+    joblib.dump(model, model_name)
+    return model
 
 
 def predict(napi, model):
     logging.info("reading prediction data")
-    napi.download_dataset(f"{DATA_VERSION}/live_universe.parquet")
-    live_universe = pd.read_parquet(f"{DATA_VERSION}/live_universe.parquet")
-
-    # TODO: implement get_features and predict the target
-    # logging.info("generating predictions")
-    # predictions = model.predict(get_features(live_universe))
-    # predictions = pd.DataFrame(
-    #     predictions, columns=["prediction"], index=predict_data.index
-    # )
-    # return predictions
-
-    # just return the latest target for now
-    napi.download_dataset(f"{DATA_VERSION}/train_targets.parquet")
-    target = pd.read_parquet(f"{DATA_VERSION}/train_targets.parquet")
-    return (
-        target[target.date == target.date.max()]
-        .drop(columns=["date"])
-        .rename(columns={"target": "prediction"})
-        .set_index("symbol")
+    napi.download_dataset(f"{DATA_VERSION}/live.parquet")
+    predict_data = pd.read_parquet(f"{DATA_VERSION}/live.parquet")
+    feature_cols = [col for col in predict_data.columns if col.startswith("feature_")]
+
+    logging.info("generating predictions")
+    predictions = model.predict(predict_data[feature_cols])
+    predictions = pd.DataFrame(
+        predictions, columns=["prediction"], index=predict_data.index
     )
+    return predictions
 
 
 def submit(predictions, predict_output_path="predictions.csv", model_id=None):
     logging.info("writing predictions to file and submitting")
     include_index = predictions.index.name is not None
     predictions.to_csv(predict_output_path, index=include_index)
-    napi.upload_predictions(
-        predict_output_path, model_id=model_id, tournament=TOURNAMENT
-    )
+    napi.upload_predictions(predict_output_path, model_id=model_id)
 
 
 if __name__ == "__main__":

numerai/examples/crypto-python3/requirements.txt

@@ -1,5 +1,6 @@
-numerapi==2.14.0
-pandas==2.1.3
-pyarrow==7.0.0
-joblib==1.2.0
-lightgbm==3.3.1
+numerapi==2.22.0
+pandas==2.3.3
+pyarrow==18.1.0
+joblib==1.5.3
+lightgbm==4.5.0
+scikit-learn==1.6.1

numerai/examples/signals-python3/Dockerfile

@@ -1,5 +1,5 @@
 # Provides us a working Python 3 environment.
-FROM python:3.9
+FROM python:3.13
 
 # These are docker arguments that `numerai node deploy/test` will always pass into docker.
 # They are then set in your environment so that numerapi can access them when uploading submissions.

numerai/examples/signals-python3/predict.py

@@ -11,7 +11,7 @@
 logging.basicConfig(filename="log.txt", filemode="a")
 
 TOURNAMENT = 11
-DATA_VERSION = "signals/v1.0"
+DATA_VERSION = "signals/v2.0"
 TARGET_COL = "target"
 TRAINED_MODEL_PREFIX = "./trained_model"
 
@@ -23,7 +23,7 @@
 # these are set by the docker image that you deploy after training,
 # but you can also set them manually above for local testing
 MODEL_ID = os.getenv("MODEL_ID", DEFAULT_MODEL_ID)
-napi = numerapi.NumerAPI(
+napi = numerapi.SignalsAPI(
     public_id=os.getenv("NUMERAI_PUBLIC_ID", DEFAULT_PUBLIC_ID),
     secret_key=os.getenv("NUMERAI_SECRET_KEY", DEFAULT_SECRET_KEY),
 )
@@ -93,9 +93,7 @@ def submit(predictions, predict_output_path="predictions.csv", model_id=None):
     include_index = predictions.index.name is not None
     predictions.to_csv(predict_output_path, index=include_index)
     print(predictions)
-    napi.upload_predictions(
-        predict_output_path, model_id=model_id, tournament=TOURNAMENT
-    )
+    napi.upload_predictions(predict_output_path, model_id=model_id)
 
 
 if __name__ == "__main__":

numerai/examples/signals-python3/requirements.txt

@@ -1,5 +1,6 @@
-numerapi==2.14.0
-pandas==2.1.3
-pyarrow==7.0.0
-joblib==1.2.0
-lightgbm==3.3.1
+numerapi==2.22.0
+pandas==2.3.3
+pyarrow==18.1.0
+joblib==1.5.3
+lightgbm==4.5.0
+scikit-learn==1.6.1

numerai/examples/tournament-python3/Dockerfile

@@ -1,5 +1,5 @@
 # Provides us a working Python 3 environment.
-FROM python:3.9
+FROM python:3.13
 
 # These are docker arguments that `numerai node deploy/test` will always pass into docker.
 # They are then set in your environment so that numerapi can access them when uploading submissions.

numerai/examples/tournament-python3/predict.py

@@ -11,10 +11,9 @@
 logging.basicConfig(filename="log.txt", filemode="a")
 
 TOURNAMENT = 8
-DATA_VERSION = "v4.1"
+DATA_VERSION = "v5.2"
 ERA_COL = "era"
-DATA_TYPE_COL = "data_type"
-TARGET_COL = "target_nomi_v4_20"
+TARGET_COL = "target"
 TRAINED_MODEL_PREFIX = "./trained_model"
 
 DEFAULT_MODEL_ID = None
@@ -35,11 +34,7 @@ def get_features(napi):
     napi.download_dataset(f"{DATA_VERSION}/features.json")
     with open(f"{DATA_VERSION}/features.json", "r") as f:
         feature_metadata = json.load(f)
-    return feature_metadata["feature_sets"]["small"] + [
-        ERA_COL,
-        DATA_TYPE_COL,
-        TARGET_COL,
-    ]
+    return feature_metadata["feature_sets"]["small"]
 
 
 def train(napi, model_id, force_training=False):
@@ -56,7 +51,8 @@ def train(napi, model_id, force_training=False):
     logging.info("reading training data")
     napi.download_dataset(f"{DATA_VERSION}/train.parquet")
     train_data = pd.read_parquet(
-        f"{DATA_VERSION}/train.parquet", columns=get_features(napi)
+        f"{DATA_VERSION}/train.parquet",
+        columns=[ERA_COL] + get_features(napi) + [TARGET_COL],
     )
 
     # This will take a few minutes 🍵
@@ -97,9 +93,7 @@ def submit(predictions, predict_output_path="predictions.csv", model_id=None):
     logging.info("writing predictions to file and submitting")
     include_index = predictions.index.name is not None
     predictions.to_csv(predict_output_path, index=include_index)
-    napi.upload_predictions(
-        predict_output_path, model_id=model_id, tournament=TOURNAMENT
-    )
+    napi.upload_predictions(predict_output_path, model_id=model_id)
 
 
 if __name__ == "__main__":