| |
| """网格实测：MIN_DF × MAX_VOCAB × 混合权重，对 12 条意图求准确率。只为找可行组合，不写文档。""" |
| import itertools |
| import os |
| import sys |
| import time |
|
|
| ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__))) |
| sys.path.insert(0, ROOT) |
| sys.path.insert(0, os.path.join(ROOT, "scripts")) |
|
|
| import bench_semvec as B |
| import holoaeon.semvec as S |
|
|
| k = B.build() |
| docs = B.corpus_texts(k) |
|
|
| print("%-8s %-6s %-6s %-8s %-10s %-10s %-10s" % ("maxvocab", "mindf", "vocab", "fit_ms", "base", "sem", "hybrid")) |
| for maxv, mindf in itertools.product([300, 800, 1200, 2000], [1, 2]): |
| S.MAX_VOCAB = maxv |
| S.MIN_DF = mindf |
| t0 = time.perf_counter() |
| idx = S.SemanticIndex() |
| idx.fit(docs) |
| fit_ms = (time.perf_counter() - t0) * 1000.0 |
| base_ok = sem_ok = hyb_ok = 0 |
| for q, expect in B.INTENTS: |
| r = B.route(k, idx, q, 0.5) |
| base_ok += r["base"]["skill"] == expect |
| sem_ok += r["sem"]["skill"] == expect |
| hyb_ok += r["hybrid"]["skill"] == expect |
| print("%-8d %-6d %-6d %-8.1f %-10s %-10s %-10s" |
| % (maxv, mindf, len(idx.vocab), fit_ms, |
| "%d/12" % base_ok, "%d/12" % sem_ok, "%d/12" % hyb_ok)) |
|
|
