frogml-cli 0.0.1__py3-none-any.whl

frogml_cli/commands/models/deployments/deploy/realtime/ui.py

@@ -0,0 +1,202 @@
+from typing import Optional
+
+import click
+from frogml.core.tools.logger.logger import (
+    get_frogml_logger,
+    set_frogml_logger_stdout_verbosity_level,
+)
+
+from frogml_cli.commands.models.deployments.deploy._logic.deploy_config import (
+    DeployConfig,
+    PurchaseOption,
+)
+from frogml_cli.commands.models.deployments.deploy._logic.deployment_response_handler import (
+    client_deployment,
+)
+from frogml_cli.commands.models.deployments.deploy._logic.local_deployment import (
+    local_deploy,
+)
+from frogml_cli.commands.models.deployments.deploy.realtime._logic.deploy_executor import (
+    RealtimeDeployExecutor,
+)
+from frogml_cli.inner.tools.cli_tools import DeprecatedOption, FrogMLCommand
+from frogml_cli.inner.tools.config_handler import config_handler
+from frogml_cli.tools.const import GPU_TYPES
+
+logger = get_frogml_logger()
+
+
+@click.command("realtime", help="Deploy realtime model.", cls=FrogMLCommand)
+@click.option(
+    "--build-id",
+    help="Build ID, If not specified latest successful build will deployed.",
+    type=str,
+)
+@click.option("--model-id", help="Model ID", type=str, required=True)
+@click.option(
+    "--timeout",
+    type=int,
+    help="Inference request timeout in MS, (e.g. 2000).",
+)
+@click.option(
+    "--server-workers",
+    type=int,
+    help="Number of workers of http server. (e.g. 4).",
+)
+@click.option(
+    "--daemon-mode/--no-daemon-mode",
+    type=bool,
+    default=True,
+    help="Configure Gunicorn daemon mode. Default is True",
+)
+@click.option(
+    "--pods",
+    "--replicas",
+    type=int,
+    help="Number of replicas to deploy.",
+    deprecated=["--pods"],
+    preferred="--replicas",
+    cls=DeprecatedOption,
+)
+@click.option(
+    "--cpus",
+    type=float,
+    help="Number of CPU cores, can be fraction of a core (e.g. 0.4).",
+)
+@click.option(
+    "--memory",
+    type=int,
+    help="Memory amount in Mib (e.g. 256).",
+)
+@click.option(
+    "--gpu-type",
+    metavar="NAME",
+    required=False,
+    help=f"Type of GPU to use on the deployments ({', '.join([x for x in GPU_TYPES])}).",
+    type=click.STRING,
+)
+@click.option(
+    "--gpu-amount",
+    metavar="NAME",
+    required=False,
+    type=int,
+    help="Amount of GPU to use on the deployments.",
+)
+@click.option(
+    "--iam-role-arn",
+    type=str,
+    help="Custom IAM Role ARN.",
+)
+@click.option(
+    "--service-account-key-secret-name",
+    type=str,
+    help="Custom service account for Gcp.",
+)
+@click.option(
+    "--max-batch-size",
+    type=int,
+    help="Max batch size in prediction. (defaults to 0 which is dynamic)",
+)
+@click.option(
+    "--variation-name",
+    required=False,
+    type=str,
+    help="The variation name",
+)
+@click.option(
+    "--sync",
+    is_flag=True,
+    default=False,
+    help="Waiting for deployments to be ready",
+)
+@click.option(
+    "--local",
+    is_flag=True,
+    default=False,
+    help="Deploy the model container locally, for development purposes only",
+)
+@click.option(
+    "-v",
+    "--verbose",
+    count=True,
+    help="Log verbosity level - v: INFO, vv: DEBUG [default: WARNING]",
+)
+@click.option(
+    "-f",
+    "--from-file",
+    help="Deploy by run_config file, Command arguments will overwrite any run_config.",
+    required=False,
+    type=click.Path(exists=True, resolve_path=True, dir_okay=False),
+)
+@click.option(
+    "--out-conf",
+    help="Extract deploy conf from commands arguments, the commands will not run it wil only output valid yaml "
+    "structure",
+    default=False,
+    is_flag=True,
+)
+@click.option(
+    "--purchase-option",
+    required=False,
+    type=click.Choice([i.value for i in PurchaseOption], case_sensitive=False),
+    help="Indicate the instance deployments type, whether it is spot/ondemand. Default is spot",
+)
+@click.option(
+    "--deployment-timeout",
+    type=int,
+    help="The number of seconds the deployments can be in progress before it is considered as failed. This is useful "
+    "in cases where a deployments has a high number of pods to replace, scarce resources (gpu), or large build "
+    "size. Default is 1800 seconds (30 minutes).",
+)
+@click.option(
+    "-E",
+    "--env-vars",
+    required=False,
+    metavar="NAME=VALUE",
+    multiple=True,
+    help="A parameter for the deployments, of the form -E name=value",
+)
+@click.option(
+    "--instance",
+    required=False,
+    type=str,
+    help="The instance size to deploy on - 'small', 'medium', 'large', etc...",
+    default=None,
+)
+@click.option(
+    "--protected",
+    "variation_protected_state",
+    required=False,
+    is_flag=True,
+    default=False,
+    help="Whether the deployment variation is protected. Default is false",
+)
+def realtime(
+    verbose: bool,
+    from_file: Optional[str],
+    out_conf: bool,
+    sync: bool,
+    local: bool,
+    **kwargs,
+):
+    set_frogml_logger_stdout_verbosity_level(verbose + 1)
+    deploy_realtime(from_file, out_conf, sync, local, **kwargs)
+
+
+def deploy_realtime(
+    from_file: Optional[str], out_conf: bool, sync: bool, local: bool, **kwargs
+):
+    config: DeployConfig = config_handler(
+        config=DeployConfig,
+        from_file=from_file,
+        out_conf=out_conf,
+        sections=("realtime", "autoscaling"),
+        **kwargs,
+    )
+
+    if local:
+        local_deploy(config)
+
+    elif not out_conf:
+        deploy_executor = RealtimeDeployExecutor(config)
+        client_deployment(deploy=deploy_executor, sync=sync)

frogml_cli/commands/models/deployments/deploy/streaming/_logic/deploy_executor.py

@@ -0,0 +1,24 @@
+from frogml._proto.qwak.deployment.deployment_pb2 import KubeDeploymentType
+from frogml._proto.qwak.deployment.deployment_service_pb2 import DeployModelResponse
+
+from frogml_cli.commands.models.deployments.deploy._logic.base_deploy_executor import (
+    BaseDeployExecutor,
+)
+from frogml_cli.commands.models.deployments.deploy._logic.deployment_message_helpers import (
+    get_env_to_deployment_message,
+)
+
+
+class StreamDeployExecutor(BaseDeployExecutor):
+    def deploy(self) -> DeployModelResponse:
+        env_deployment_messages = get_env_to_deployment_message(
+            self.config,
+            KubeDeploymentType.STREAM,
+            self.ecosystem_client,
+            self.instance_template_client,
+        )
+        return self.deploy_client.deploy_model(
+            model_id=self.config.model_id,
+            build_id=self.config.build_id,
+            env_deployment_messages=env_deployment_messages,
+        )

frogml_cli/commands/models/deployments/deploy/streaming/_logic/serving_strategy_mapper.py

@@ -0,0 +1,38 @@
+from __future__ import annotations
+
+from frogml._proto.qwak.deployment.deployment_pb2 import (
+    KafkaConfig,
+    ServingStrategy,
+    StreamConfig,
+)
+
+from frogml_cli.commands.models.deployments.deploy._logic.deploy_config import (
+    DeployConfig,
+)
+
+DEFAULT_VARIATION_NAME = "default"
+
+
+def create_streaming_serving_strategy_from_deploy_config(
+    deploy_config: DeployConfig,
+) -> ServingStrategy:
+    return ServingStrategy(
+        stream_config=StreamConfig(
+            kafka=KafkaConfig(
+                consumer=KafkaConfig.Consumer(
+                    bootstrap_server=deploy_config.stream.kafka.consumer.bootstrap_servers,
+                    topic=deploy_config.stream.kafka.consumer.topic,
+                    group=deploy_config.stream.kafka.consumer.group,
+                    timeout=deploy_config.stream.kafka.consumer.timeout,
+                    auto_offset_type=deploy_config.stream.kafka.consumer.auto_offset_reset.upper(),
+                    max_batch_size=deploy_config.stream.kafka.consumer.max_batch_size,
+                    max_poll_latency=deploy_config.stream.kafka.consumer.max_poll_latency,
+                ),
+                producer=KafkaConfig.Producer(
+                    bootstrap_server=deploy_config.stream.kafka.producer.bootstrap_servers,
+                    topic=deploy_config.stream.kafka.producer.topic,
+                    compression_type=deploy_config.stream.kafka.producer.compression_type.upper(),
+                ),
+            )
+        )
+    )

frogml_cli/commands/models/deployments/deploy/streaming/ui.py

@@ -0,0 +1,206 @@
+import click
+from frogml.core.tools.logger.logger import (
+    get_frogml_logger,
+    set_frogml_logger_stdout_verbosity_level,
+)
+
+from frogml_cli.commands.models.deployments.deploy._logic.deploy_config import (
+    AutoOffsetReset,
+    CompressionTypes,
+    DeployConfig,
+    PurchaseOption,
+)
+from frogml_cli.commands.models.deployments.deploy._logic.deployment_response_handler import (
+    client_deployment,
+)
+from frogml_cli.commands.models.deployments.deploy.streaming._logic.deploy_executor import (
+    StreamDeployExecutor,
+)
+from frogml_cli.inner.tools.cli_tools import DeprecatedOption, FrogMLCommand
+from frogml_cli.inner.tools.config_handler import config_handler
+
+logger = get_frogml_logger()
+PENDING_STATUSES = [
+    "INITIATING_DEPLOYMENT",
+    "PENDING_DEPLOYMENT",
+    "INITIATING_UNDEPLOYMENT",
+    "PENDING_UNDEPLOYMENT",
+]
+DEPLOYED_STATUSES = ["SUCCESSFUL_DEPLOYMENT"]
+UNDEPLOYED_STATUSES = ["SUCCESSFUL_UNDEPLOYMENT"]
+FAILED_STATUSES = [
+    "INVALID_DEPLOYMENT",
+    "FAILED_INITIATING_DEPLOYMENT",
+    "FAILED_DEPLOYMENT",
+    "FAILED_UNDEPLOYMENT",
+]
+
+
+@click.command("stream", help="Deploy stream model.", cls=FrogMLCommand)
+@click.option(
+    "--build-id",
+    help="Build ID, If not specified latest successful build will deployed.",
+    type=str,
+)
+@click.option("--model-id", help="Model ID", type=str, required=True)
+@click.option(
+    "--bootstrap-server",
+    envvar="BOOTSTRAP_SERVERS",
+    help="Kafka consumer/producer bootstrap server.",
+    multiple=True,
+    type=str,
+)
+@click.option(
+    "--consumer-bootstrap-server",
+    envvar="CONSUMER_BOOTSTRAP_SERVER",
+    help="Kafka consumer bootstrap server.",
+    multiple=True,
+    type=str,
+)
+@click.option(
+    "--consumer-topic",
+    envvar="CONSUMER_TOPIC",
+    help="Kafka consumer topic.",
+    type=str,
+)
+@click.option(
+    "--consumer-group",
+    envvar="CONSUMER_GROUP",
+    help="Kafka consumer group.",
+    type=str,
+)
+@click.option(
+    "--consumer-auto-offset-reset",
+    envvar="CONSUMER_AUTO_OFFSET_RESET",
+    type=click.Choice([i.value for i in AutoOffsetReset], case_sensitive=False),
+    help="Kafka consumer auto offset reset (Latest/Earliest).",
+)
+@click.option(
+    "--consumer-timeout",
+    envvar="CONSUMER_TIMEOUT",
+    help="Kafka consumer polling timeout. (The timeout should be in range of kafka admin configuration "
+    "'group.min.session.timeout.ms' and 'group.max.session.timeout.ms'.)",
+    type=int,
+)
+@click.option(
+    "--consumer-max-batch-size",
+    envvar="CONSUMER_MAX_BATCH_SIZE",
+    help="The maximum number of records returned in a single call to poll().",
+    type=int,
+)
+@click.option(
+    "--consumer-max-poll-latency",
+    envvar="CONSUMER_MAX_POLL_LATENCY",
+    help="The maximum delay between invocations of poll() when using consumer group management.",
+    type=float,
+)
+@click.option(
+    "--producer-bootstrap-server",
+    envvar="PRODUCER_BOOTSTRAP_SERVER",
+    help="Kafka producer bootstrap server.",
+    multiple=True,
+    type=str,
+)
+@click.option(
+    "--producer-topic",
+    envvar="PRODUCER_TOPIC",
+    help="Kafka producer topic.",
+    type=str,
+)
+@click.option(
+    "--producer-compression-type",
+    envvar="PRODUCER_COMPRESSION_TYPE",
+    type=click.Choice([i.value for i in CompressionTypes], case_sensitive=False),
+    help="Kafka producer compression type.",
+)
+@click.option(
+    "--pods",
+    "--replicas",
+    type=int,
+    help="Number of replicas to deploy.",
+    deprecated=["--pods"],
+    preferred="--replicas",
+    cls=DeprecatedOption,
+)
+@click.option(
+    "--cpus",
+    type=float,
+    help="Number of CPU cores, can be fraction of a core (e.g. 0.4).",
+)
+@click.option(
+    "--memory",
+    type=int,
+    help="Memory amount in Mib (e.g. 256).",
+)
+@click.option(
+    "--iam-role-arn",
+    type=str,
+    help="Custom IAM Role ARN.",
+)
+@click.option(
+    "--sync",
+    is_flag=True,
+    default=False,
+    help="Waiting for deployments to be ready",
+)
+@click.option(
+    "--local",
+    is_flag=True,
+    default=False,
+    help="Run deployments locally.",
+)
+@click.option(
+    "-v",
+    "--verbose",
+    count=True,
+    help="Log verbosity level - v: INFO, vv: DEBUG [default: WARNING]",
+)
+@click.option(
+    "-f",
+    "--from-file",
+    help="Deploy by run_config file, Command arguments will overwrite any run_config.",
+    required=False,
+    type=click.Path(exists=True, resolve_path=True, dir_okay=False),
+)
+@click.option(
+    "--out-conf",
+    help="Extract deploy conf from commands arguments, the commands will not run it wil only output valid yaml "
+    "structure",
+    default=False,
+    is_flag=True,
+)
+@click.option(
+    "--purchase-option",
+    required=False,
+    type=click.Choice([i.value for i in PurchaseOption], case_sensitive=False),
+    help="Indicate the instance deployments type, whether it is spot/ondemand.  Default is spot",
+)
+@click.option(
+    "-E",
+    "--env-vars",
+    required=False,
+    metavar="NAME=VALUE",
+    multiple=True,
+    help="A parameter for the deployments, of the form -E name=value",
+)
+@click.option(
+    "--instance",
+    required=False,
+    type=str,
+    help="The instance size to deploy on - 'small', 'medium', 'large', etc...",
+    default=None,
+)
+def stream(
+    verbose: bool, from_file: str, out_conf: bool, local: bool, sync: bool, **kwargs
+):
+    set_frogml_logger_stdout_verbosity_level(verbose + 1)
+    config: DeployConfig = config_handler(
+        config=DeployConfig,
+        from_file=from_file,
+        out_conf=out_conf,
+        sections=("stream",),
+        **kwargs,
+    )
+    if not out_conf:
+        deploy_executor = StreamDeployExecutor(config)
+        client_deployment(deploy=deploy_executor, sync=sync)