Understanding Sentence Transformers negative sampling code

Question

I want to train a recommender model in PyTorch. I have only positive pairs (e.g., user A bought product P), and I wanted to implement in-batch negative mining, since that technique is referenced so often in recommender system research papers. I came across this Sentence Transformers PyTorch code that I'm hoping to adapt. In the doc string it says "For each a_i, it uses all other p_j as negative samples", however in the code it seems only the first item in the batch is assigned negative pairs. I'm specifically looking at the section:

embeddings_a = reps[0]
embeddings_b = torch.cat(reps[1:])

Below is the complete code. Am I misunderstanding something?

from typing import Dict, Iterable

import torch
from torch import Tensor, nn

from sentence_transformers import util
from sentence_transformers.SentenceTransformer import SentenceTransformer


class MultipleNegativesRankingLoss(nn.Module):
    def __init__(self, model: SentenceTransformer, scale: float = 20.0, similarity_fct=util.cos_sim):
        """
        This loss expects as input a batch consisting of sentence pairs ``(a_1, p_1), (a_2, p_2)..., (a_n, p_n)``
        where we assume that ``(a_i, p_i)`` are a positive pair and ``(a_i, p_j)`` for ``i != j`` a negative pair.

        For each ``a_i``, it uses all other ``p_j`` as negative samples, i.e., for ``a_i``, we have 1 positive example
        (``p_i``) and ``n-1`` negative examples (``p_j``). It then minimizes the negative log-likehood for softmax
        normalized scores.

        This loss function works great to train embeddings for retrieval setups where you have positive pairs
        (e.g. (query, relevant_doc)) as it will sample in each batch ``n-1`` negative docs randomly.

        The performance usually increases with increasing batch sizes.

        You can also provide one or multiple hard negatives per anchor-positive pair by structering the data like this:
        ``(a_1, p_1, n_1), (a_2, p_2, n_2)``. Then, ``n_1`` is a hard negative for ``(a_1, p_1)``. The loss will use for
        the pair ``(a_i, p_i)`` all ``p_j`` for ``j != i`` and all ``n_j`` as negatives.

        Args:
            model: SentenceTransformer model
            scale: Output of similarity function is multiplied by scale
                value
            similarity_fct: similarity function between sentence
                embeddings. By default, cos_sim. Can also be set to dot
                product (and then set scale to 1)

        """
        super(MultipleNegativesRankingLoss, self).__init__()
        self.model = model
        self.scale = scale
        self.similarity_fct = similarity_fct
        self.cross_entropy_loss = nn.CrossEntropyLoss()

    def forward(self, sentence_features: Iterable[Dict[str, Tensor]], labels: Tensor):
        reps = [self.model(sentence_feature)["sentence_embedding"] for sentence_feature in sentence_features]
        embeddings_a = reps[0]
        embeddings_b = torch.cat(reps[1:])

        scores = self.similarity_fct(embeddings_a, embeddings_b) * self.scale
        # Example a[i] should match with b[i]
        range_labels = torch.arange(0, scores.size(0), device=scores.device)
        return self.cross_entropy_loss(scores, range_labels)