File size: 1,502 Bytes
c5e4524
 
 
bb343a5
 
 
 
9c6f23d
c5e4524
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch


def create_dense_embeddings(query, model, instruction=None):
    if instruction == None:
        dense_emb = model.encode([query]).tolist()
    else:
        dense_emb = model.encode([[instruction, query]]).tolist()
    return dense_emb


def create_sparse_embeddings(query, model, tokenizer):
    device = "cuda" if torch.cuda.is_available() else "cpu"
    inputs = tokenizer(query, return_tensors="pt").to(device)

    with torch.no_grad():
        logits = model(**inputs).logits

    inter = torch.log1p(torch.relu(logits[0]))
    token_max = torch.max(inter, dim=0)  # sum over input tokens
    nz_tokens = torch.where(token_max.values > 0)[0]
    nz_weights = token_max.values[nz_tokens]

    order = torch.sort(nz_weights, descending=True)
    nz_weights = nz_weights[order[1]]
    nz_tokens = nz_tokens[order[1]]
    return {
        "indices": nz_tokens.cpu().numpy().tolist(),
        "values": nz_weights.cpu().numpy().tolist(),
    }


def hybrid_score_norm(dense, sparse, alpha: float):
    """Hybrid score using a convex combination

    alpha * dense + (1 - alpha) * sparse

    Args:
        dense: Array of floats representing
        sparse: a dict of `indices` and `values`
        alpha: scale between 0 and 1
    """
    if alpha < 0 or alpha > 1:
        raise ValueError("Alpha must be between 0 and 1")
    hs = {
        "indices": sparse["indices"],
        "values": [v * (1 - alpha) for v in sparse["values"]],
    }
    return [v * alpha for v in dense], hs