numerai-cli 1.0.3__tar.gz → 1.1.0__tar.gz

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: numerai-cli
-Version: 1.0.3
+Version: 1.1.0
 Summary: A library for deploying Numer.ai Prediction Nodes.
 Home-page: https://github.com/numerai/numerai-cli
 Author: Numer.ai
@@ -11,6 +11,7 @@ Description: # numerai-cli
         [![PyPI](https://img.shields.io/pypi/v/numerai-cli.svg?color=brightgreen)](https://pypi.org/project/numerai-cli/)
         
         Welcome to the Numerai CLI for the [Numerai Tournament](https://docs.numer.ai/tournament/learn).
+        Numerai CLI (aka Compute Heavy) is a tool to help you set up cloud automation for your Numerai predictions.
         This README is designed to have EVERYTHING you need to setup and maintain a Numerai Compute Heavy node.
         
         This CLI runs on your local computer to configure a Numerai Prediction Node in the cloud.

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/README.md

@@ -3,6 +3,7 @@
 [![PyPI](https://img.shields.io/pypi/v/numerai-cli.svg?color=brightgreen)](https://pypi.org/project/numerai-cli/)
 
 Welcome to the Numerai CLI for the [Numerai Tournament](https://docs.numer.ai/tournament/learn).
+Numerai CLI (aka Compute Heavy) is a tool to help you set up cloud automation for your Numerai predictions.
 This README is designed to have EVERYTHING you need to setup and maintain a Numerai Compute Heavy node.
 
 This CLI runs on your local computer to configure a Numerai Prediction Node in the cloud.

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/numerai/cli/constants.py

@@ -1,10 +1,9 @@
 """Constants for Numerai CLI"""
+
 import os
 import json
 from pathlib import Path
 
-TOURNAMENT_NUMERAI = 8
-TOURNAMENT_SIGNALS = 11
 PACKAGE_PATH = os.path.dirname(__file__)
 CONFIG_PATH = os.path.join(str(Path.home()), ".numerai")
 KEYS_PATH = os.path.join(CONFIG_PATH, ".keys")
@@ -65,11 +64,6 @@ DEFAULT_SETTINGS = {
 }
 
 CONSTANTS_STR = f"""Default values (not your configured node values):
-
----Tournament Numbers---
-TOURNAMENT_NUMERAI: {TOURNAMENT_NUMERAI}
-TOURNAMENT_SIGNALS: {TOURNAMENT_SIGNALS}
-
 ---Paths---
 PACKAGE_PATH: {PACKAGE_PATH}
 CONFIG_PATH: {CONFIG_PATH}

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/numerai/cli/node/__init__.py

@@ -1,4 +1,5 @@
 """Init for node"""
+
 import json
 import logging
 import click
@@ -15,22 +16,23 @@ from numerai.cli.util.keys import get_numerai_keys
 # Setting azure's logging level "ERROR" to avoid spamming the terminal
 
 
-def get_models(signals):
-    if signals:
-        tournament = TOURNAMENT_SIGNALS
-        name_prefix = "signals"
-    else:
-        tournament = TOURNAMENT_NUMERAI
-        name_prefix = "numerai"
+def tournaments_dict():
+    napi = base_api.Api()
+    tournaments = napi.raw_query('query { tournaments { name tournament } }')
+    return {t["tournament"]: t["name"] for t in tournaments["data"]["tournaments"]}
+
+
+def get_models(tournament):
     napi = base_api.Api(*get_numerai_keys())
     models = napi.get_models(tournament)
-
+    tournaments = napi.raw_query('query { tournaments { name tournament } }')
+    name_prefix = tournaments_dict()[tournament]
     model_dict = {}
     for model_name, model_id in models.items():
         model_dict[model_name] = {
             "id": model_id,
             "name": f"{name_prefix}-{model_name}",
-            "is_signals": signals,
+            "tournament": tournament,
         }
     return model_dict
 
@@ -43,16 +45,18 @@ def get_models(signals):
     type=str,
     prompt=True,
     help="The name of one of your models to configure the Prediction Node for."
-    "It defaults to the first model returned from your account.",
+    " It defaults to the first model returned from your account.",
 )
 @click.option(
-    "--signals",
-    "-s",
-    is_flag=True,
-    help="Target a signals model with this name. Defaults to false.",
+    "--tournament",
+    "-t",
+    default=8,
+    help="Target a specific tournament number."
+    " Defaults to Numerai Tournament/Classic."
+    f" Available tournaments: {json.dumps(tournaments_dict(), indent=2)}",
 )
 @click.pass_context
-def node(ctx, verbose, model_name, signals):
+def node(ctx, verbose, model_name, tournament):
     """
     Commands to manage and test Prediction Nodes.
     """
@@ -68,7 +72,7 @@ def node(ctx, verbose, model_name, signals):
     else:
         logger.setLevel(logging.ERROR)
 
-    models = get_models(signals)
+    models = get_models(tournament)
 
     try:
         ctx.ensure_object(dict)

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/numerai/cli/node/config.py

@@ -1,4 +1,5 @@
 """Config command for Numerai CLI"""
+
 import json
 import os
 import click
@@ -94,6 +95,12 @@ from numerai.cli.util.keys import get_provider_keys, get_numerai_keys, load_or_i
     help="Forces your webhook to register with Numerai. "
     "Use in conjunction with options that prevent webhook auto-registering.",
 )
+@click.option(
+    "--volume",
+    "-v",
+    type=int,
+    help="Specify additional block storage in GB. Currently only supported in AWS.",
+)
 @click.pass_context
 def config(
     ctx,
@@ -107,6 +114,7 @@ def config(
     cron,
     timeout_minutes,
     register_webhook,
+    volume,
 ):
     """
     Uses Terraform to create a full Numerai Compute cluster in your desired provider.
@@ -262,12 +270,12 @@ def config(
         provider_registry_conf = create_gcp_registry(provider, verbose=verbose)
         node_conf.update(provider_registry_conf)
         registry_parts = node_conf["registry_id"].split("/")
-        node_conf[
-            "artifact_registry_login_url"
-        ] = f"https://{registry_parts[3]}-docker.pkg.dev/"
-        node_conf[
-            "docker_repo"
-        ] = f"{registry_parts[3]}-docker.pkg.dev/{registry_parts[1]}/numerai-container-registry/{node}:latest"
+        node_conf["artifact_registry_login_url"] = (
+            f"https://{registry_parts[3]}-docker.pkg.dev/"
+        )
+        node_conf["docker_repo"] = (
+            f"{registry_parts[3]}-docker.pkg.dev/{registry_parts[1]}/numerai-container-registry/{node}:latest"
+        )
         docker.login(node_conf, verbose)
         try:
             docker.manifest_inspect(node_conf["docker_repo"], verbose)
@@ -276,6 +284,9 @@ def config(
             docker.tag("hello-world:linux", node_conf["docker_repo"], verbose)
             docker.push(node_conf["docker_repo"], verbose)
         nodes_config[node] = node_conf
+    elif provider == "aws":
+        if volume is not None:
+            node_conf["volume"] = volume
 
     store_config(NODES_PATH, nodes_config)
 

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/numerai/cli/node/test.py

@@ -61,7 +61,7 @@ def test(ctx, local, command, verbose):
     ctx.ensure_object(dict)
     model = ctx.obj["model"]
     node = model["name"]
-    is_signals = model["is_signals"]
+    tournament = model["tournament"]
     node_config = load_or_init_nodes(node)
     provider = node_config["provider"]
 
@@ -143,7 +143,6 @@ def test(ctx, local, command, verbose):
         variables={"modelId": node_config["model_id"]},
         authorization=True,
     )
-    tournament = TOURNAMENT_SIGNALS if is_signals else TOURNAMENT_NUMERAI
     curr_round = api.get_current_round(tournament)
     latest_subs = sorted(
         filter(
@@ -338,7 +337,7 @@ def get_recent_task_status_aws(cluster_arn, ecs_client, node, trigger_id):
             "red",
         )
     elif matched_task["lastStatus"] in stopped_codes:
-        return matched_task, True, "Job execution succeeded!\r", "green"
+        return matched_task, True, "Job execution finished!\r", "green"
     elif matched_task["lastStatus"] in pending_codes:
         return matched_task, False, "Waiting for job to start...", "yellow"
     elif matched_task["lastStatus"] in running_codes:

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/numerai/cli/util/docker.py

@@ -166,9 +166,16 @@ def build(node_config, node, verbose):
     build_arg_str += f" --build-arg NODE={node}"
 
     cmd = (
-        f'docker build --platform=linux/amd64 -t {node_config["docker_repo"]}' f"{build_arg_str} -f {path}/Dockerfile ."
+        f'docker build --platform=linux/amd64 --load -t {node_config["docker_repo"]}'
+        f"{build_arg_str} -f {path}/Dockerfile ."
     )
-    execute(cmd, verbose)
+    stdout, stderr = execute(cmd, verbose)
+    if "unknown flag: --load" in (stdout.decode() + stderr.decode()):
+        click.secho(
+            "Docker version too old, please upgrade to at least 18.09",
+            fg="red",
+        )
+        exit(1)
 
 
 def run(node_config, verbose, command=""):
@@ -181,7 +188,9 @@ def login(node_config, verbose):
         username, password = login_aws()
         login_url = node_config['docker_repo']
     elif node_config["provider"] == PROVIDER_AZURE:
-        username, password = login_azure(node_config["registry_rg_name"], node_config["registry_name"])
+        username, password = login_azure(
+            node_config["registry_rg_name"], node_config["registry_name"]
+        )
         login_url = node_config['docker_repo']
     elif node_config["provider"] == PROVIDER_GCP:
         username, password = login_gcp()
@@ -214,17 +223,23 @@ def login_aws():
     )
 
     token = ecr_client.get_authorization_token()  # TODO: use registryIds
-    username, password = base64.b64decode(token["authorizationData"][0]["authorizationToken"]).decode().split(":")
+    username, password = (
+        base64.b64decode(token["authorizationData"][0]["authorizationToken"])
+        .decode()
+        .split(":")
+    )
 
     return username, password
 
 
 def login_azure(resource_group_name, registry_name):
     azure_subs_id, azure_client, azure_tenant, azure_secret = get_azure_keys()
-    credentials = ClientSecretCredential(client_id=azure_client, tenant_id=azure_tenant, client_secret=azure_secret)
-    username_password = ContainerRegistryManagementClient(credentials, azure_subs_id).registries.list_credentials(
-        resource_group_name, registry_name
+    credentials = ClientSecretCredential(
+        client_id=azure_client, tenant_id=azure_tenant, client_secret=azure_secret
     )
+    username_password = ContainerRegistryManagementClient(
+        credentials, azure_subs_id
+    ).registries.list_credentials(resource_group_name, registry_name)
     username = username_password.username
     password = username_password.passwords[0].value
     return username, password
@@ -288,24 +303,32 @@ def cleanup_aws(docker_repo):
 
     docker_repo_name = docker_repo.split("/")[-1]
 
-    resp = ecr_client.list_images(repositoryName=docker_repo_name, filter={"tagStatus": "UNTAGGED"})
+    resp = ecr_client.list_images(
+        repositoryName=docker_repo_name, filter={"tagStatus": "UNTAGGED"}
+    )
 
     imageIds = resp["imageIds"]
     if len(imageIds) == 0:
         return []
 
-    resp = ecr_client.batch_delete_image(repositoryName=docker_repo_name, imageIds=imageIds)
+    resp = ecr_client.batch_delete_image(
+        repositoryName=docker_repo_name, imageIds=imageIds
+    )
 
     return resp["imageIds"]
 
 
 def cleanup_azure(node_config):
     _, azure_client, azure_tenant, azure_secret = get_azure_keys()
-    credentials = ClientSecretCredential(client_id=azure_client, tenant_id=azure_tenant, client_secret=azure_secret)
+    credentials = ClientSecretCredential(
+        client_id=azure_client, tenant_id=azure_tenant, client_secret=azure_secret
+    )
     acr_client = ContainerRegistryClient(node_config["acr_login_server"], credentials)
     docker_repo = node_config["docker_repo"]
     node_repo_name = [
-        repo_name for repo_name in acr_client.list_repository_names() if repo_name == docker_repo.split("/")[-1]
+        repo_name
+        for repo_name in acr_client.list_repository_names()
+        if repo_name == docker_repo.split("/")[-1]
     ][0]
 
     # get all manifests, ordered by last update time
@@ -318,7 +341,9 @@ def cleanup_azure(node_config):
     # Remove all but the latest manifest
     removed_manifests = []
     for manifest in manifest_list[1:]:
-        acr_client.update_manifest_properties(node_repo_name, manifest.digest, can_write=True, can_delete=True)
+        acr_client.update_manifest_properties(
+            node_repo_name, manifest.digest, can_write=True, can_delete=True
+        )
         removed_manifests.append(manifest.digest)
         acr_client.delete_manifest(node_repo_name, manifest.digest)
     return removed_manifests

{numerai-cli-1.0.3/numerai/examples/tournament-python3 → numerai-cli-1.1.0/numerai/examples/crypto-python3}/Dockerfile

@@ -4,9 +4,9 @@ FROM python:3.9
 # 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.
 # You can also access them from your script like so:
-    # import os
-    # public_id = os.environ["NUMERAI_PUBLIC_ID"]
-    # secret_key = os.environ["NUMERAI_SECRET_KEY"]
+# import os
+# public_id = os.environ["NUMERAI_PUBLIC_ID"]
+# secret_key = os.environ["NUMERAI_SECRET_KEY"]
 ARG NUMERAI_PUBLIC_ID
 ENV NUMERAI_PUBLIC_ID=$NUMERAI_PUBLIC_ID
 

numerai-cli-1.1.0/numerai/examples/crypto-python3/predict.py

@@ -0,0 +1,105 @@
+""" Sample tournament model in python 3 """
+
+import os
+import json
+import logging
+import joblib
+import numerapi
+import pandas as pd
+import lightgbm as lgbm
+
+logging.basicConfig(filename="log.txt", filemode="a")
+
+TOURNAMENT = 12
+DATA_VERSION = "crypto/v1.0"
+TRAINED_MODEL_PREFIX = "./trained_model"
+
+DEFAULT_MODEL_ID = None
+DEFAULT_PUBLIC_ID = None
+DEFAULT_SECRET_KEY = None
+
+# Read model id and initialize API client with api keys
+# 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(
+    public_id=os.getenv("NUMERAI_PUBLIC_ID", DEFAULT_PUBLIC_ID),
+    secret_key=os.getenv("NUMERAI_SECRET_KEY", DEFAULT_SECRET_KEY),
+)
+
+
+def train(napi, model_id, force_training=False):
+    model_name = TRAINED_MODEL_PREFIX
+    if model_id:
+        model_name += f"_{model_id}"
+
+    # load a model if we have a trained model already and we aren't forcing a training session
+    if os.path.exists(model_name) and not force_training:
+        logging.info("loading existing trained model")
+        model = joblib.load(model_name)
+        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
+
+
+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")
+    )
+
+
+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
+    )
+
+
+if __name__ == "__main__":
+    trained_model = train(napi, MODEL_ID)
+    predictions = predict(napi, trained_model)
+    submit(predictions, model_id=MODEL_ID)

numerai-cli-1.1.0/numerai/examples/crypto-python3/requirements.txt

@@ -0,0 +1,5 @@
+numerapi==2.14.0
+pandas==2.1.3
+pyarrow==7.0.0
+joblib==1.2.0
+lightgbm==3.3.1

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/numerai/examples/signals-python3/Dockerfile

@@ -1,12 +1,12 @@
 # Provides us a working Python 3 environment.
-FROM python:3.7
+FROM python:3.9
 
 # 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.
 # You can also access them from your script like so:
-    # import os
-    # public_id = os.environ["NUMERAI_PUBLIC_ID"]
-    # secret_key = os.environ["NUMERAI_SECRET_KEY"]
+# import os
+# public_id = os.environ["NUMERAI_PUBLIC_ID"]
+# secret_key = os.environ["NUMERAI_SECRET_KEY"]
 ARG NUMERAI_PUBLIC_ID
 ENV NUMERAI_PUBLIC_ID=$NUMERAI_PUBLIC_ID
 

numerai-cli-1.1.0/numerai/examples/signals-python3/predict.py

@@ -0,0 +1,104 @@
+""" Sample tournament model in python 3 """
+
+import os
+import json
+import logging
+import joblib
+import numerapi
+import pandas as pd
+import lightgbm as lgbm
+
+logging.basicConfig(filename="log.txt", filemode="a")
+
+TOURNAMENT = 11
+DATA_VERSION = "signals/v1.0"
+TARGET_COL = "target"
+TRAINED_MODEL_PREFIX = "./trained_model"
+
+DEFAULT_MODEL_ID = None
+DEFAULT_PUBLIC_ID = None
+DEFAULT_SECRET_KEY = None
+
+# Read model id and initialize API client with api keys
+# 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(
+    public_id=os.getenv("NUMERAI_PUBLIC_ID", DEFAULT_PUBLIC_ID),
+    secret_key=os.getenv("NUMERAI_SECRET_KEY", DEFAULT_SECRET_KEY),
+)
+
+
+def train(napi, model_id, force_training=False):
+    model_name = TRAINED_MODEL_PREFIX
+    if model_id:
+        model_name += f"_{model_id}"
+
+    # load a model if we have a trained model already and we aren't forcing a training session
+    if os.path.exists(model_name) and not force_training:
+        logging.info("loading existing trained model")
+        model = joblib.load(model_name)
+        return model
+
+    logging.info("reading training data")
+    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_')
+        and col not in ("feature_country", "feature_exchange_code")
+    ]
+
+    # 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
+
+
+def predict(napi, model):
+    logging.info("reading prediction data")
+    napi.download_dataset(f"{DATA_VERSION}/live.parquet")
+    predict_data = pd.read_parquet(f"{DATA_VERSION}/live.parquet").set_index(
+        'numerai_ticker'
+    )
+    feature_cols = [
+        col
+        for col in predict_data.columns
+        if col.startswith('feature_')
+        and col not in ("feature_country", "feature_exchange_code")
+    ]
+    print(predict_data)
+
+    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)
+    print(predictions)
+    napi.upload_predictions(
+        predict_output_path, model_id=model_id, tournament=TOURNAMENT
+    )
+
+
+if __name__ == "__main__":
+    trained_model = train(napi, MODEL_ID)
+    predictions = predict(napi, trained_model)
+    submit(predictions, model_id=MODEL_ID)

numerai-cli-1.1.0/numerai/examples/signals-python3/requirements.txt

@@ -0,0 +1,5 @@
+numerapi==2.14.0
+pandas==2.1.3
+pyarrow==7.0.0
+joblib==1.2.0
+lightgbm==3.3.1

numerai-cli-1.1.0/numerai/examples/tournament-python3/Dockerfile

@@ -0,0 +1,35 @@
+# Provides us a working Python 3 environment.
+FROM python:3.9
+
+# 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.
+# You can also access them from your script like so:
+# import os
+# public_id = os.environ["NUMERAI_PUBLIC_ID"]
+# secret_key = os.environ["NUMERAI_SECRET_KEY"]
+ARG NUMERAI_PUBLIC_ID
+ENV NUMERAI_PUBLIC_ID=$NUMERAI_PUBLIC_ID
+
+ARG NUMERAI_SECRET_KEY
+ENV NUMERAI_SECRET_KEY=$NUMERAI_SECRET_KEY
+
+ARG MODEL_ID
+ENV MODEL_ID=$MODEL_ID
+
+ARG SRC_PATH
+ENV SRC_PATH=$SRC_PATH
+
+# We then add the requirements.txt file, and pip install every requirement from it.
+# The `ADD [source] [destination]` command will take a file from the source directory on your computer
+# and copy it over to the destination directory in the Docker container.
+ADD $SRC_PATH/requirements.txt .
+RUN pip install -r requirements.txt --no-cache-dir
+
+# Now, add everything in the source code directory.
+# (including your code, compiled files, serialized models, everything...)
+ADD $SRC_PATH .
+
+# This sets the default command to run your docker container.
+# It runs by default in the cloud and when running `numerai node test`.
+# This is overridden when using `numerai node test --command [COMMAND]`.
+CMD [ "python", "./predict.py" ]

{numerai-cli-1.0.3 → numerai-cli-1.1.0}/numerai/examples/tournament-python3/predict.py

@@ -6,10 +6,11 @@ import logging
 import joblib
 import numerapi
 import pandas as pd
-from sklearn.linear_model import LinearRegression
+import lightgbm as lgbm
 
 logging.basicConfig(filename="log.txt", filemode="a")
 
+TOURNAMENT = 8
 DATA_VERSION = "v4.1"
 ERA_COL = "era"
 DATA_TYPE_COL = "data_type"
@@ -52,18 +53,24 @@ def train(napi, model_id, force_training=False):
         model = joblib.load(model_name)
         return model
 
-    model = LinearRegression()
-
     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)
     )
 
+    # 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(
-        X=train_data.filter(like="feature_", axis="columns"),
-        y=train_data[TARGET_COL],
+        train_data.filter(like="feature_", axis="columns"),
+        train_data[TARGET_COL],
     )
 
     logging.info("saving model")
@@ -88,8 +95,11 @@ def predict(napi, model):
 
 def submit(predictions, predict_output_path="predictions.csv", model_id=None):
     logging.info("writing predictions to file and submitting")
-    predictions.to_csv(predict_output_path)
-    napi.upload_predictions(predict_output_path, model_id=model_id)
+    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
+    )
 
 
 if __name__ == "__main__":

numerai-cli-1.1.0/numerai/examples/tournament-python3/requirements.txt

@@ -0,0 +1,5 @@
+numerapi==2.14.0
+pandas==2.1.3
+pyarrow==7.0.0
+joblib==1.2.0
+lightgbm==3.3.1

numerai-cli-1.1.0/numerai/examples/tournament-python3/train.py

@@ -0,0 +1,10 @@
+""" An extra entry point specifically for training. Used when running locally """
+
+import predict
+
+train_data_path, predict_data_path, predict_output_path = predict.download_data()
+
+model_id = predict.MODEL_ID
+model_type = predict.MODEL
+
+predict.train(train_data_path, model_id, model_type, force_training=True)