Views
No views yet
| Model | Initialization | MARCO Dev | Encoder Path |
|---|---|---|---|
| aggretriever-distilbert | distilbert-base-uncased | 34.1 | castorini/aggretriever-distilbert |
| aggretriever-cocondenser | Luyu/co-condenser-marco | 36.2 | castorini/aggretriever-cocondenser |
1from pyserini.encode._aggretriever import AggretrieverQueryEncoder
2from pyserini.encode._aggretriever import AggretrieverDocumentEncoder
3
4model_name = '/store/scratch/s269lin/experiments/aggretriever/hf_model/aggretriever-cocondenser'
5query_encoder = AggretrieverQueryEncoder(model_name, device='cpu')
6context_encoder = AggretrieverDocumentEncoder(model_name, device='cpu')
7
8query = ["Where was Marie Curie born?"]
9contexts = [
10 "Maria Sklodowska, later known as Marie Curie, was born on November 7, 1867.",
11 "Born in Paris on 15 May 1859, Pierre Curie was the son of Eugène Curie, a doctor of French Catholic origin from Alsace."
12]
13
14# Compute embeddings: take the last-layer hidden state of the [CLS] token
15query_emb = query_encoder.encode(query)
16ctx_emb = context_encoder.encode(contexts)
17# Compute similarity scores using dot product
18score1 = query_emb @ ctx_emb[0] # 45.56658
19score2 = query_emb @ ctx_emb[1] # 45.81762