1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64
|
// Copyright 2021 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#ifndef COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_SCHEDULER_MODEL_EXECUTION_SCHEDULER_H_
#define COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_SCHEDULER_MODEL_EXECUTION_SCHEDULER_H_
#include "components/segmentation_platform/public/proto/segmentation_platform.pb.h"
namespace segmentation_platform {
namespace proto {
class SegmentInfo;
} // namespace proto
struct ModelExecutionResult;
using proto::SegmentId;
// Central class responsible for scheduling model execution. Determines which
// models are eligible for execution based on various criteria e.g. cached
// results, TTL etc. Invoked from multiple classes such as segment
// selector or periodic jobs.
class ModelExecutionScheduler {
public:
// An observer to be notified about the model execution and results.
class Observer {
public:
// Called whenever a model execution completes.
virtual void OnModelExecutionCompleted(SegmentId segment_id) = 0;
};
virtual ~ModelExecutionScheduler() = default;
// Called whenever a new or updated model is available. Must be a valid
// SegmentInfo with valid metadata and features.
virtual void OnNewModelInfoReady(const proto::SegmentInfo& segment_info) = 0;
// Central method to determine which all models to execute. Called in response
// to segmentation requests from clients, or periodic tasks. Can be called
// repeatedly. Loops through all the models and decides whether to execute a
// model or not. A segment is considered eligible to run based on
// - if it has no results, or
// - if the results have expired or
// - if the results weren't computed too recently, and |expired_only| is false
// - if all the required signals have been collected for sufficient duration.
virtual void RequestModelExecutionForEligibleSegments(bool expired_only) = 0;
// Runs model execution for a particular segment.
virtual void RequestModelExecution(
const proto::SegmentInfo& segment_info) = 0;
// Called after model execution completes. If the execution was successful,
// saves the results to the DB, and notifies observers. If the execution was
// unsuccessful, deletes the result from the DB.
// TODO(shaktisahu): Do we want to store that failure reason in the DB
// instead? We might treat different failures differently next time.
virtual void OnModelExecutionCompleted(
const proto::SegmentInfo& segment_info,
std::unique_ptr<ModelExecutionResult> result) = 0;
};
} // namespace segmentation_platform
#endif // COMPONENTS_SEGMENTATION_PLATFORM_INTERNAL_SCHEDULER_MODEL_EXECUTION_SCHEDULER_H_
|