tensorflow_similarity/models/contrastive_model.py [1048:1135]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        if not self._index.is_calibrated:
            raise ValueError("Uncalibrated model: run model.calibration()")

        cal_metric = self._index.get_calibration_metric()

        # get embeddings
        if verbose:
            print("|-Computing embeddings")
        predictions = self.predict(x)

        results: DefaultDict[
            str, Dict[str, Union[str, np.ndarray]]
        ] = defaultdict(dict)

        if verbose:
            pb = tqdm(
                total=len(self._index.cutpoints), desc="Evaluating cutpoints"
            )

        for cp_name, cp_data in self._index.cutpoints.items():
            # create a metric that match at the requested k and threshold
            distance_threshold = float(cp_data["distance"])
            metric = make_classification_metric(cal_metric.name)
            metrics = copy(extra_metrics)
            metrics.append(metric)

            res: Dict[str, Union[str, np.ndarray]] = {}
            res.update(
                self._index.evaluate_classification(
                    predictions,
                    y,
                    [distance_threshold],
                    metrics=metrics,
                    matcher=matcher,
                    k=k,
                )
            )
            res["distance"] = tf.constant([distance_threshold])
            res["name"] = cp_name
            results[cp_name] = res
            if verbose:
                pb.update()

        if verbose:
            pb.close()

        if verbose:
            headers = ["name", cal_metric.name]
            for i in results["optimal"].keys():
                if i not in headers:
                    headers.append(str(i))
            rows = []
            for data in results.values():
                rows.append([data[v] for v in headers])
            print("\n [Summary]\n")
            print(tabulate(rows, headers=headers))

        return results

    def reset_index(self):
        "Reinitialize the index"
        self._index.reset()

    def index_size(self) -> int:
        "Return the index size"
        return self._index.size()

    def load_index(self, filepath: str):
        """Load Index data from a checkpoint and initialize underlying
        structure with the reloaded data.

        Args:
            path: Directory where the checkpoint is located.
            verbose: Be verbose. Defaults to 1.
        """

        index_path = Path(filepath) / "index"
        self._index = Indexer.load(index_path)

    def save_index(self, filepath, compression=True):
        """Save the index to disk

        Args:
            path: directory where to save the index
            compression: Store index data compressed. Defaults to True.
        """
        index_path = Path(filepath) / "index"
        self._index.save(index_path, compression=compression)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



tensorflow_similarity/models/similarity_model.py [678:765]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        if not self._index.is_calibrated:
            raise ValueError("Uncalibrated model: run model.calibration()")

        cal_metric = self._index.get_calibration_metric()

        # get embeddings
        if verbose:
            print("|-Computing embeddings")
        predictions = self.predict(x)

        results: DefaultDict[
            str, Dict[str, Union[str, np.ndarray]]
        ] = defaultdict(dict)

        if verbose:
            pb = tqdm(
                total=len(self._index.cutpoints), desc="Evaluating cutpoints"
            )

        for cp_name, cp_data in self._index.cutpoints.items():
            # create a metric that match at the requested k and threshold
            distance_threshold = float(cp_data["distance"])
            metric = make_classification_metric(cal_metric.name)
            metrics = copy(extra_metrics)
            metrics.append(metric)

            res: Dict[str, Union[str, np.ndarray]] = {}
            res.update(
                self._index.evaluate_classification(
                    predictions,
                    y,
                    [distance_threshold],
                    metrics=metrics,
                    matcher=matcher,
                    k=k,
                )
            )
            res["distance"] = tf.constant([distance_threshold])
            res["name"] = cp_name
            results[cp_name] = res
            if verbose:
                pb.update()

        if verbose:
            pb.close()

        if verbose:
            headers = ["name", cal_metric.name]
            for i in results["optimal"].keys():
                if i not in headers:
                    headers.append(str(i))
            rows = []
            for data in results.values():
                rows.append([data[v] for v in headers])
            print("\n [Summary]\n")
            print(tabulate(rows, headers=headers))

        return results

    def reset_index(self):
        "Reinitialize the index"
        self._index.reset()

    def index_size(self) -> int:
        "Return the index size"
        return self._index.size()

    def load_index(self, filepath: str):
        """Load Index data from a checkpoint and initialize underlying
        structure with the reloaded data.

        Args:
            path: Directory where the checkpoint is located.
            verbose: Be verbose. Defaults to 1.
        """

        index_path = Path(filepath) / "index"
        self._index = Indexer.load(index_path)

    def save_index(self, filepath, compression=True):
        """Save the index to disk

        Args:
            path: directory where to save the index
            compression: Store index data compressed. Defaults to True.
        """
        index_path = Path(filepath) / "index"
        self._index.save(index_path, compression=compression)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -