def start(args):
api = None
try:
ctx = Context(s3_path=args.context,
cache_dir=args.cache_dir,
workload_id=args.workload_id)
api = ctx.apis_id_map[args.api]
local_cache["api"] = api
local_cache["ctx"] = ctx
if api.get("onnx") is None:
raise CortexException(api["name"], "onnx key not configured")
_, prefix = ctx.storage.deconstruct_s3_path(api["onnx"]["model"])
model_path = os.path.join(args.model_dir, os.path.basename(prefix))
if api["onnx"].get("request_handler") is not None:
local_cache["request_handler"] = ctx.get_request_handler_impl(
api["name"], args.project_dir)
request_handler = local_cache.get("request_handler")
if request_handler is not None and util.has_function(
request_handler, "pre_inference"):
cx_logger().info(
"using pre_inference request handler provided in {}".format(
api["onnx"]["request_handler"]))
else:
cx_logger().info("pre_inference request handler not found")
if request_handler is not None and util.has_function(
request_handler, "post_inference"):
cx_logger().info(
"using post_inference request handler provided in {}".format(
api["onnx"]["request_handler"]))
else:
cx_logger().info("post_inference request handler not found")
sess = rt.InferenceSession(model_path)
local_cache["sess"] = sess
local_cache["input_metadata"] = sess.get_inputs()
cx_logger().info("input_metadata: {}".format(
truncate(extract_signature(local_cache["input_metadata"]))))
local_cache["output_metadata"] = sess.get_outputs()
cx_logger().info("output_metadata: {}".format(
truncate(extract_signature(local_cache["output_metadata"]))))
except Exception as e:
cx_logger().exception("failed to start api")
sys.exit(1)
if api.get("tracker") is not None and api["tracker"].get(
"model_type") == "classification":
try:
local_cache["class_set"] = api_utils.get_classes(ctx, api["name"])
except Exception as e:
cx_logger().warn(
"an error occurred while attempting to load classes",
exc_info=True)
cx_logger().info("API is ready")
serve(app, listen="*:{}".format(args.port))
def start(args):
api = None
try:
ctx = Context(s3_path=args.context,
cache_dir=args.cache_dir,
workload_id=args.workload_id)
api = ctx.apis_id_map[args.api]
local_cache["api"] = api
local_cache["ctx"] = ctx
if api.get("request_handler_impl_key") is not None:
package.install_packages(ctx.python_packages, ctx.storage)
local_cache["request_handler"] = ctx.get_request_handler_impl(
api["name"])
model_cache_path = os.path.join(args.model_dir, args.api)
if not os.path.exists(model_cache_path):
ctx.storage.download_file_external(api["model"], model_cache_path)
sess = rt.InferenceSession(model_cache_path)
local_cache["sess"] = sess
local_cache["input_metadata"] = sess.get_inputs()
local_cache["output_metadata"] = sess.get_outputs()
except CortexException as e:
e.wrap("error")
logger.error(str(e))
if api is not None:
logger.exception(
"An error occured starting the api, see `cx logs -v api {}` for more details"
.format(api["name"]))
sys.exit(1)
serve(app, listen="*:{}".format(args.port))
def start(args):
api = None
try:
ctx = Context(s3_path=args.context,
cache_dir=args.cache_dir,
workload_id=args.workload_id)
api = ctx.apis_id_map[args.api]
local_cache["api"] = api
local_cache["ctx"] = ctx
if api.get("request_handler") is not None:
local_cache["request_handler"] = ctx.get_request_handler_impl(
api["name"], args.project_dir)
except Exception as e:
logger.exception("failed to start api")
sys.exit(1)
try:
validate_model_dir(args.model_dir)
except Exception as e:
logger.exception("failed to validate model")
sys.exit(1)
if api.get("tracker") is not None and api["tracker"].get(
"model_type") == "classification":
try:
local_cache["class_set"] = api_utils.get_classes(ctx, api["name"])
except Exception as e:
logger.warn("an error occurred while attempting to load classes",
exc_info=True)
channel = grpc.insecure_channel("localhost:" + str(args.tf_serve_port))
local_cache["stub"] = prediction_service_pb2_grpc.PredictionServiceStub(
channel)
# wait a bit for tf serving to start before querying metadata
limit = 60
for i in range(limit):
try:
local_cache["metadata"] = run_get_model_metadata()
break
except Exception as e:
if i > 6:
logger.warn(
"unable to read model metadata - model is still loading. Retrying..."
)
if i == limit - 1:
logger.exception("retry limit exceeded")
sys.exit(1)
time.sleep(5)
logger.info("model_signature: {}".format(
extract_signature(
local_cache["metadata"]["signatureDef"],
local_cache["api"]["tf_serving"]["signature_key"],
)))
serve(app, listen="*:{}".format(args.port))
def start(args):
api = None
try:
ctx = Context(s3_path=args.context,
cache_dir=args.cache_dir,
workload_id=args.workload_id)
api = ctx.apis_id_map[args.api]
local_cache["api"] = api
local_cache["ctx"] = ctx
_, prefix = ctx.storage.deconstruct_s3_path(api["model"])
model_path = os.path.join(args.model_dir, os.path.basename(prefix))
if api.get("request_handler") is not None:
local_cache["request_handler"] = ctx.get_request_handler_impl(
api["name"], args.project_dir)
sess = rt.InferenceSession(model_path)
local_cache["sess"] = sess
local_cache["input_metadata"] = sess.get_inputs()
logger.info("input_metadata: {}".format(
truncate(extract_signature(local_cache["input_metadata"]))))
local_cache["output_metadata"] = sess.get_outputs()
logger.info("output_metadata: {}".format(
truncate(extract_signature(local_cache["output_metadata"]))))
except Exception as e:
logger.exception("failed to start api")
sys.exit(1)
if api.get("tracker") is not None and api["tracker"].get(
"model_type") == "classification":
try:
local_cache["class_set"] = api_utils.get_classes(ctx, api["name"])
except Exception as e:
logger.warn("an error occurred while attempting to load classes",
exc_info=True)
serve(app, listen="*:{}".format(args.port))
def start(args):
ctx = Context(s3_path=args.context,
cache_dir=args.cache_dir,
workload_id=args.workload_id)
api = ctx.apis_id_map[args.api]
local_cache["api"] = api
local_cache["ctx"] = ctx
if api.get("request_handler_impl_key") is not None:
package.install_packages(ctx.python_packages, ctx.storage)
local_cache["request_handler"] = ctx.get_request_handler_impl(
api["name"])
model_cache_path = os.path.join(args.model_dir, args.api)
if not os.path.exists(model_cache_path):
ctx.storage.download_file_external(api["model"], model_cache_path)
sess = rt.InferenceSession(model_cache_path)
local_cache["sess"] = sess
local_cache["input_metadata"] = sess.get_inputs()
local_cache["output_metadata"] = sess.get_outputs()
logger.info("Serving model: {}".format(
util.remove_resource_ref(api["model"])))
serve(app, listen="*:{}".format(args.port))
def start(args):
api = None
try:
ctx = Context(s3_path=args.context,
cache_dir=args.cache_dir,
workload_id=args.workload_id)
api = ctx.apis_id_map[args.api]
local_cache["api"] = api
local_cache["ctx"] = ctx
if api.get("tensorflow") is None:
raise CortexException(api["name"], "tensorflow key not configured")
if api["tensorflow"].get("request_handler") is not None:
cx_logger().info("loading the request handler from {}".format(
api["tensorflow"]["request_handler"]))
local_cache["request_handler"] = ctx.get_request_handler_impl(
api["name"], args.project_dir)
request_handler = local_cache.get("request_handler")
if request_handler is not None and util.has_function(
request_handler, "pre_inference"):
cx_logger().info(
"using pre_inference request handler defined in {}".format(
api["tensorflow"]["request_handler"]))
else:
cx_logger().info("pre_inference request handler not defined")
if request_handler is not None and util.has_function(
request_handler, "post_inference"):
cx_logger().info(
"using post_inference request handler defined in {}".format(
api["tensorflow"]["request_handler"]))
else:
cx_logger().info("post_inference request handler not defined")
except Exception as e:
cx_logger().exception("failed to start api")
sys.exit(1)
try:
validate_model_dir(args.model_dir)
except Exception as e:
cx_logger().exception("failed to validate model")
sys.exit(1)
if api.get("tracker") is not None and api["tracker"].get(
"model_type") == "classification":
try:
local_cache["class_set"] = api_utils.get_classes(ctx, api["name"])
except Exception as e:
cx_logger().warn(
"an error occurred while attempting to load classes",
exc_info=True)
channel = grpc.insecure_channel("localhost:" + str(args.tf_serve_port))
local_cache["stub"] = prediction_service_pb2_grpc.PredictionServiceStub(
channel)
# wait a bit for tf serving to start before querying metadata
limit = 60
for i in range(limit):
try:
local_cache["model_metadata"] = run_get_model_metadata()
break
except Exception as e:
if i > 6:
cx_logger().warn(
"unable to read model metadata - model is still loading. Retrying..."
)
if i == limit - 1:
cx_logger().exception("retry limit exceeded")
sys.exit(1)
time.sleep(5)
signature_key, parsed_signature = extract_signature(
local_cache["model_metadata"]["signatureDef"],
api["tensorflow"]["signature_key"])
local_cache["signature_key"] = signature_key
local_cache["parsed_signature"] = parsed_signature
cx_logger().info("model_signature: {}".format(
local_cache["parsed_signature"]))
cx_logger().info("{} API is live".format(api["name"]))
serve(app, listen="*:{}".format(args.port))
def start(args):
ctx = Context(s3_path=args.context,
cache_dir=args.cache_dir,
workload_id=args.workload_id)
api = ctx.apis_id_map[args.api]
local_cache["api"] = api
local_cache["ctx"] = ctx
try:
if api.get("request_handler_impl_key") is not None:
local_cache["request_handler"] = ctx.get_request_handler_impl(
api["name"])
if not util.is_resource_ref(api["model"]):
if api.get("request_handler") is not None:
package.install_packages(ctx.python_packages, ctx.storage)
if not os.path.isdir(args.model_dir):
ctx.storage.download_and_unzip_external(
api["model"], args.model_dir)
else:
package.install_packages(ctx.python_packages, ctx.storage)
model_name = util.get_resource_ref(api["model"])
model = ctx.models[model_name]
estimator = ctx.estimators[model["estimator"]]
local_cache["model"] = model
local_cache["estimator"] = estimator
local_cache["target_col"] = ctx.columns[util.get_resource_ref(
model["target_column"])]
local_cache["target_col_type"] = ctx.get_inferred_column_type(
util.get_resource_ref(model["target_column"]))
log_level = "DEBUG"
if ctx.environment is not None and ctx.environment.get(
"log_level") is not None:
log_level = ctx.environment["log_level"].get(
"tensorflow", "DEBUG")
tf_lib.set_logging_verbosity(log_level)
if not os.path.isdir(args.model_dir):
ctx.storage.download_and_unzip(model["key"], args.model_dir)
for column_name in ctx.extract_column_names(
[model["input"], model["target_column"]]):
if ctx.is_transformed_column(column_name):
trans_impl, _ = ctx.get_transformer_impl(column_name)
local_cache["trans_impls"][column_name] = trans_impl
transformed_column = ctx.transformed_columns[column_name]
# cache aggregate values
for resource_name in util.extract_resource_refs(
transformed_column["input"]):
if resource_name in ctx.aggregates:
ctx.get_obj(ctx.aggregates[resource_name]["key"])
local_cache["required_inputs"] = tf_lib.get_base_input_columns(
model["name"], ctx)
if util.is_dict(model["input"]) and model["input"].get(
"target_vocab") is not None:
local_cache["target_vocab_populated"] = ctx.populate_values(
model["input"]["target_vocab"], None, False)
except CortexException as e:
e.wrap("error")
logger.error(str(e))
logger.exception(
"An error occurred, see `cortex logs -v api {}` for more details.".
format(api["name"]))
sys.exit(1)
except Exception as e:
logger.exception(
"An error occurred, see `cortex logs -v api {}` for more details.".
format(api["name"]))
sys.exit(1)
try:
validate_model_dir(args.model_dir)
except Exception as e:
logger.exception(e)
sys.exit(1)
channel = grpc.insecure_channel("localhost:" + str(args.tf_serve_port))
local_cache["stub"] = prediction_service_pb2_grpc.PredictionServiceStub(
channel)
# wait a bit for tf serving to start before querying metadata
limit = 300
for i in range(limit):
try:
local_cache["metadata"] = run_get_model_metadata()
break
except Exception as e:
if i == limit - 1:
logger.exception(
"An error occurred, see `cortex logs -v api {}` for more details."
.format(api["name"]))
sys.exit(1)
time.sleep(1)
serve(app, listen="*:{}".format(args.port))
