Merge 6ccc5ad into 21c9c2b

tests/test_benchmarks.py

@@ -4,14 +4,16 @@
 from vecto.benchmarks.similarity import Similarity
 from vecto.benchmarks.analogy import *
 from vecto.embeddings import load_from_dir
+from vecto.benchmarks.fetch_benchmarks import fetch_benchmarks
+from os import path
+from shutil import rmtree
 
 
 path_similarity_dataset = "./tests/data/benchmarks/similarity/"
 path_analogy_dataset = "./tests/data/benchmarks/analogy/"
 
-class Tests(unittest.TestCase):
-
 
+class Tests(unittest.TestCase):
     def test_similarity(self):
         embs = load_from_dir("tests/data/embeddings/text/plain_with_file_header")
         similarity = Similarity()
@@ -23,17 +25,19 @@ def test_similarity(self):
         print(result)
 
     def test_analogy(self):
-        embs = load_from_dir("./tests/data/embeddings/text/plain_with_file_header")
-
+        embs = load_from_dir("tests/data/embeddings/text/plain_with_file_header")
         analogy = LinearOffset()
         result = analogy.get_result(embs, path_analogy_dataset)
         print(result)
+
         analogy = PairDistance()
         result = analogy.get_result(embs, path_analogy_dataset)
         print(result)
+
         analogy = ThreeCosMul()
         result = analogy.get_result(embs, path_analogy_dataset)
         print(result)
+
         analogy = ThreeCosMul2()
         result = analogy.get_result(embs, path_analogy_dataset)
         print(result)
@@ -47,6 +51,16 @@ def test_analogy(self):
         result = analogy.get_result(embs, path_analogy_dataset)
         print(result)
 
+    def test_fetcher(self):
+        if path.isdir('./tests/data/benchmarks_test'):
+            rmtree('./tests/data/benchmarks_test')
+        fetch_benchmarks('./tests/data/benchmarks_test')
+        embs = load_from_dir("tests/data/embeddings/text/plain_with_file_header")
+        similarity = Similarity()
+        path_similarity_dataset = "./tests/data/benchmarks_test/benchmarks/similarity/en/"
+        result = similarity.get_result(embs, path_similarity_dataset)
+
+
         # big embs and dataset test
         # embs = load_from_dir("/home/bofang/Documents/embeddings/negative_sampling/fair/")
         # result = analogy.get_result(embs, "/home/bofang/Downloads/BATS_3.0_small")

vecto/benchmarks/analogy/analogy.py

@@ -383,6 +383,8 @@ def run(self, embs, path_dataset):
         results = []
         for root, dirnames, filenames in os.walk(dir_tests):
             for filename in fnmatch.filter(sorted(filenames), '*'):
+                if filename.endswith('json'):
+                    continue
                 print(filename)
                 pairs = self.get_pairs(os.path.join(root, filename))
                 # print(pairs)

vecto/benchmarks/fetch_benchmarks.py

@@ -3,10 +3,9 @@
 from os import path
 
 path_to_repo = 'https://github.com/vecto-ai/benchmarks.git'
-path_to_local_dir = path.join('data', 'benchmarks')
 
 
-def fetch_benchmarks():
+def fetch_benchmarks(path_to_local_dir=path.join('data', 'benchmarks')):
     try:
         Repo.clone_from('https://github.com/vecto-ai/benchmarks.git', path_to_local_dir)
     except GitCommandError:

vecto/benchmarks/similarity/similarity.py

@@ -51,6 +51,8 @@ def evaluate(self, embs, data):
     def run(self, embs, path_dataset):
         results = []
         for file in os.listdir(path_dataset):
+            if file.endswith('json'):
+                continue
             testset = self.read_test_set(os.path.join(path_dataset, file))
 
             out = dict()