jsonl.py 文件源码

def __iter__(self):
        """ Return an iterator to the data. Get the value (tensor) for self.key
        from each object and yield batches of these tensors
        """
        start = 0
        while start < self.num_entries:
            end = min(self.num_entries, start + self.chunk_size)

            # linecache line numbering starts at 1
            batch = [
                json.loads(linecache.getline(self.source, i + 1).strip())[self.key]
                for i in self.indices[start:end]
            ]

            yield batch
            start = end

    ###########################################################################