bug fix for punc and umap

shixian.shi

2024-01-23 ae4dceecf03c643be770e3274a070078e45fc9c7

bug fix for punc and umap

 funasr/models/campplus/cluster_backend.py

@@ -119,6 +119,7 @@
        self.metric = metric

    def __call__(self, X):
        from umap.umap_ import UMAP
        umap_X = umap.UMAP(
            n_neighbors=self.n_neighbors,
            min_dist=0.0,
@@ -156,6 +157,7 @@
        if X.shape[0] < 20:
            return np.zeros(X.shape[0], dtype='int')
        if X.shape[0] < 2048 or k is not None:
            # unexpected corner case
            labels = self.spectral_cluster(X, k)
        else:
            labels = self.umap_hdbscan_cluster(X)

 funasr/models/ct_transformer/model.py

@@ -336,10 +336,11 @@
                elif new_mini_sentence[-1] != "。" and new_mini_sentence[-1] != "？" and len(new_mini_sentence[-1].encode())!=1:
                    new_mini_sentence_out = new_mini_sentence + "。"
                    new_mini_sentence_punc_out = new_mini_sentence_punc[:-1] + [self.sentence_end_id]
                    if len(punctuations): punctuations[-1] = 2
                elif new_mini_sentence[-1] != "." and new_mini_sentence[-1] != "?" and len(new_mini_sentence[-1].encode())==1:
                    new_mini_sentence_out = new_mini_sentence + "."
                    new_mini_sentence_punc_out = new_mini_sentence_punc[:-1] + [self.sentence_end_id]
                    
                    if len(punctuations): punctuations[-1] = 2
            # keep a punctuations array for punc segment
            if punc_array is None:
                punc_array = punctuations
@@ -347,6 +348,5 @@
                punc_array = torch.cat([punc_array, punctuations], dim=0)
        result_i = {"key": key[0], "text": new_mini_sentence_out, "punc_array": punc_array}
        results.append(result_i)
    
        return results, meta_data


 setup.py

@@ -38,7 +38,7 @@
        # "protobuf",
        "tqdm",
        "hdbscan",
        "umap",
        "umap_learn",
        "jaconv",
        "hydra-core>=1.3.2",
    ],