| import sys,numpy as np,pandas as pd |
| sys.path.insert(0,'/mnt/data') |
| import msmarco_full_search_post as slow |
| import msmarco_full_search_fast as fast |
| tr=pd.read_csv('/mnt/data/dev.tsv',sep='\t',usecols=['query-id']); ids=[str(x) for x in np.unique(tr['query-id'].to_numpy())[:10]]; del tr |
| txt=fast.load_query_texts(ids) |
| a=slow.FullIndex(); b=fast.FullIndex() |
| for q in ids: |
| pa=a.prepare(txt[q],20); pb=b.prepare(txt[q],20) |
| for h in [0,1,5,10,20]: |
| ra=a.rank_h(pa,h,100); rb=b.rank_h(pb,h,100) |
| if ra!=rb: |
| print('MISMATCH',q,h,next((i for i,(x,y) in enumerate(zip(ra,rb)) if x!=y),None)); break |
| else: print('OK',q) |
|
|
