fix python blank space lints (facebookresearch#4560)

limqiying · facebook-github-bot · commit 8b83ebbf6271 · 2025-08-26T19:59:55.000-07:00
Summary: Pull Request resolved: facebookresearch#4560 **What:** Fixed 73 Python \E302 linting errors by adding missing blank lines before function definitions. **Why:** Python PEP 8 style guide requires 2 blank lines before top-level function and class definitions. The codebase had functions with only 1 blank line preceding them, causing linting failures. **Changes:** Added one additional blank line before function definitions across 12 files in the faiss codebase to meet the "expected 2 blank lines, found 1" requirement. **Impact:** Brings Python code into compliance with standard formatting rules without affecting functionality. Reviewed By: trang-nm-nguyen Differential Revision: D81080938 fbshipit-source-id: dd5e02c52229ca77db38b465bdaadd9bfa83801a
diff --git a/benchs/bench_all_ivf/bench_all_ivf.py b/benchs/bench_all_ivf/bench_all_ivf.py
@@ -104,6 +104,7 @@ def eval_setting(index, xq, gt, k, inter, min_time):
 # Training
 ######################################################
 
+
 def run_train(args, ds, res):
     nq, d = ds.nq, ds.d
     nb, d = ds.nq, ds.d
@@ -235,6 +236,7 @@ def run_train(args, ds, res):
 # Populating index
 ######################################################
 
+
 def run_add(args, ds, index, res):
 
     print("adding")
@@ -564,4 +566,4 @@ def aa(*args, **kwargs):
 
 
 if __name__ == "__main__":
-    main()
+    main()
diff --git a/benchs/bench_all_ivf/cmp_with_scann.py b/benchs/bench_all_ivf/cmp_with_scann.py
@@ -23,6 +23,7 @@ def eval_recalls(name, I, gt, times):
     s += "time: %.4f s (± %.4f)" % (np.mean(times), np.std(times))
     print(s)
 
+
 def eval_inters(name, I, gt, times):
     k = I.shape[1]
     s = "%-40s inter" % name
@@ -168,6 +169,7 @@ def aa(*args, **kwargs):
 # SCANN
 ###############################################################
 
+
 def scann_make_index(xb, distance_measure, scann_dir, reorder_k):
     import scann
 
diff --git a/benchs/bench_all_ivf/datasets_oss.py b/benchs/bench_all_ivf/datasets_oss.py
@@ -17,6 +17,7 @@
 
 faiss_datasets.dataset_basedir = '/checkpoint/matthijs/simsearch/'
 
+
 def sanitize(x):
     return np.ascontiguousarray(x, dtype='float32')
 
diff --git a/benchs/bench_all_ivf/parse_bench_all_ivf.py b/benchs/bench_all_ivf/parse_bench_all_ivf.py
@@ -153,6 +153,7 @@ def collect_results_for(db='deep1M', prefix="autotune."):
 
     return allres, allstats
 
+
 def extract_pareto_optimal(allres, keys, recall_idx=0, times_idx=3):
     bigtab = []
     for i, k in enumerate(keys):
@@ -183,6 +184,7 @@ def extract_pareto_optimal(allres, keys, recall_idx=0, times_idx=3):
 
     return selected_keys, ops
 
+
 def plot_subset(
     allres, allstats, selected_methods, recall_idx, times_idx=3,
     report=["overhead", "build time"]):
diff --git a/benchs/bench_for_interrupt.py b/benchs/bench_for_interrupt.py
@@ -14,6 +14,7 @@
 
 parser = argparse.ArgumentParser()
 
+
 def aa(*args, **kwargs):
     group.add_argument(*args, **kwargs)
 
diff --git a/benchs/bench_fw_codecs.py b/benchs/bench_fw_codecs.py
@@ -13,6 +13,7 @@
 
 logging.basicConfig(level=logging.INFO)
 
+
 def factory_factory(d):
     return [
         ("SQ4", None, 256 * (2 ** 10), None),
@@ -77,6 +78,7 @@ def factory_factory(d):
         if d % sub == 0
     ]
 
+
 def run_local(rp):
     bio, d, tablename, distance_metric = rp
     if tablename == "contriever":
@@ -120,6 +122,7 @@ def run_local(rp):
     benchmark.set_io(bio)
     benchmark.benchmark(result_file="result.json", train=True, reconstruct=False, knn=False, range=False)
 
+
 def run(bio, d, tablename, distance_metric):
     bio.launch_jobs(run_local, [(bio, d, tablename, distance_metric)], local=True)
 
diff --git a/benchs/bench_fw_ivf.py b/benchs/bench_fw_ivf.py
@@ -71,6 +71,7 @@ def bigann(bio):
         benchmark.set_io(bio)
         benchmark.benchmark(f"result{scale}.json", local=False, train=True, reconstruct=False, knn=True, range=False)
 
+
 def ssnpp(bio):
     benchmark = Benchmark(
         num_threads=32,
diff --git a/benchs/bench_gpu_1bn.py b/benchs/bench_gpu_1bn.py
@@ -123,11 +123,13 @@ def usage():
 # we mem-map the biggest files to avoid having them in memory all at
 # once
 
+
 def mmap_fvecs(fname):
     x = np.memmap(fname, dtype='int32', mode='r')
     d = x[0]
     return x.view('float32').reshape(-1, d + 1)[:, 1:]
 
+
 def mmap_bvecs(fname):
     x = np.memmap(fname, dtype='uint8', mode='r')
     d = x[:4].view('int32')[0]
@@ -560,6 +562,7 @@ def compute_populated_index(preproc):
 
     return gpu_index, indexall
 
+
 def compute_populated_index_2(preproc):
 
     indexall = prepare_trained_index(preproc)
diff --git a/benchs/bench_pq_tables.py b/benchs/bench_pq_tables.py
@@ -11,12 +11,12 @@
 
 os.system("grep -m1 'model name' < /proc/cpuinfo")
 
+
 def format_tab(x):
     return "\n".join("\t".join("%g" % xi for xi in row) for row in x)
 
 
 def run_bench(d, dsub, nbit=8, metric=None):
-
     M = d // dsub
     pq = faiss.ProductQuantizer(d, M, nbit)
     pq.train(faiss.randn((max(1000, pq.ksub * 50), d), 123))
diff --git a/benchs/distributed_ondisk/distributed_query_demo.py b/benchs/distributed_ondisk/distributed_query_demo.py
@@ -43,11 +43,13 @@
 sindex.set_prefetch_nthread(0)
 sindex.set_omp_num_threads(64)
 
+
 def ivecs_read(fname):
     a = np.fromfile(fname, dtype='int32')
     d = a[0]
     return a.reshape(-1, d + 1)[:, 1:].copy()
 
+
 def fvecs_read(fname):
     return ivecs_read(fname).view('float32')
 
diff --git a/benchs/distributed_ondisk/make_index_vslice.py b/benchs/distributed_ondisk/make_index_vslice.py
@@ -9,11 +9,13 @@
 import argparse
 from multiprocessing.pool import ThreadPool
 
+
 def ivecs_mmap(fname):
     a = np.memmap(fname, dtype='int32', mode='r')
     d = a[0]
     return a.reshape(-1, d + 1)[:, 1:]
 
+
 def fvecs_mmap(fname):
     return ivecs_mmap(fname).view('float32')
 
@@ -56,6 +58,7 @@ def next_or_None():
 deep1bdir = "/datasets01_101/simsearch/041218/deep1b/"
 workdir = "/checkpoint/matthijs/ondisk_distributed/"
 
+
 def main():
     parser = argparse.ArgumentParser(
         description='make index for a subset of the data')
diff --git a/benchs/distributed_ondisk/make_trained_index.py b/benchs/distributed_ondisk/make_trained_index.py
@@ -34,11 +34,13 @@
         )
     )
 
+
 def ivecs_mmap(fname):
     a = np.memmap(fname, dtype='int32', mode='r')
     d = a[0]
     return a.reshape(-1, d + 1)[:, 1:]
 
+
 def fvecs_mmap(fname):
     return ivecs_mmap(fname).view('float32')
 
diff --git a/benchs/distributed_ondisk/search_server.py b/benchs/distributed_ondisk/search_server.py
@@ -16,7 +16,6 @@
 # Server implementation
 ############################################################
 
-
 class MyServer(rpc.Server):
     """ Assign version that can be exposed via RPC """
     def __init__(self, s, index):
@@ -26,6 +25,7 @@ def __init__(self, s, index):
     def __getattr__(self, f):
         return getattr(self.index, f)
 
+
 def main():
     parser = argparse.ArgumentParser()
 
@@ -97,6 +97,7 @@ def add_batch_result(self, D, I, i0):
     def finalize(self):
         self.heaps.reorder()
 
+
 def distribute_weights(weights, nbin):
     """ assign a set of weights to a smaller set of bins to balance them """
     nw = weights.size
diff --git a/benchs/kmeans_mnist.py b/benchs/kmeans_mnist.py
@@ -18,6 +18,7 @@
 
 # Load Leon's file format
 
+
 def load_mnist(fname):
     print("load", fname)
     f = open(fname)
diff --git a/contrib/clustering.py b/contrib/clustering.py
@@ -18,9 +18,11 @@
 except ImportError:
     print("scipy not accessible, Python k-means will not work")
 
+
 def print_nop(*arg, **kwargs):
     pass
 
+
 def two_level_clustering(xt, nc1, nc2, rebalance=True, clustering_niter=25, **args):
     """
     perform 2-level clustering on a training set xt
diff --git a/contrib/evaluation.py b/contrib/evaluation.py
@@ -13,6 +13,7 @@
 ###############################################################
 # Simple functions to evaluate knn results
 
+
 def knn_intersection_measure(I1, I2):
     """ computes the intersection measure of two result tables
     """
@@ -27,6 +28,7 @@ def knn_intersection_measure(I1, I2):
 ###############################################################
 # Range search results can be compared with Precision-Recall
 
+
 def filter_range_results(lims, D, I, thresh):
     """ select a set of results """
     nq = lims.size - 1
@@ -123,6 +125,7 @@ def counts_to_PR(ngt, nres, ninter, mode="overall"):
     else:
         raise AssertionError()
 
+
 def sort_range_res_2(lims, D, I):
     """ sort 2 arrays using the first as key """
     I2 = np.empty_like(I)
diff --git a/contrib/exhaustive_search.py b/contrib/exhaustive_search.py
@@ -11,6 +11,7 @@
 
 LOG = logging.getLogger(__name__)
 
+
 def knn_ground_truth(xq, db_iterator, k, metric_type=faiss.METRIC_L2, shard=False, ngpu=-1):
     """Computes the exact KNN search results for a dataset that possibly
     does not fit in RAM but for which we have an iterator that
diff --git a/contrib/torch_utils.py b/contrib/torch_utils.py
@@ -56,6 +56,7 @@ def swig_ptr_from_FloatTensor(x):
     return faiss.cast_integer_to_float_ptr(
         x.untyped_storage().data_ptr() + x.storage_offset() * 4)
 
+
 def swig_ptr_from_BFloat16Tensor(x):
     """ gets a Faiss SWIG pointer from a pytorch tensor (on CPU or GPU) """
     assert x.is_contiguous()
@@ -92,6 +93,7 @@ def swig_ptr_from_Int64Tensor(x):
 # utilities
 ##################################################################
 
+
 @contextlib.contextmanager
 def using_stream(res, pytorch_stream=None):
     """ Creates a scoping object to make Faiss GPU use the same stream
diff --git a/demos/demo_auto_tune.py b/demos/demo_auto_tune.py
@@ -22,11 +22,13 @@
 # Small I/O functions
 #################################################################
 
+
 def ivecs_read(fname):
     a = np.fromfile(fname, dtype="int32")
     d = a[0]
     return a.reshape(-1, d + 1)[:, 1:].copy()
 
+
 def fvecs_read(fname):
     return ivecs_read(fname).view('float32')
 
diff --git a/faiss/gpu/perf/slow.py b/faiss/gpu/perf/slow.py
@@ -8,6 +8,7 @@
 import faiss
 import numpy as np
 
+
 def test_slow():
     d = 256
     index = faiss.index_cpu_to_gpu(faiss.StandardGpuResources(),
diff --git a/faiss/gpu/test/test_cagra.py b/faiss/gpu/test/test_cagra.py
@@ -10,6 +10,7 @@
 from faiss.contrib import datasets, evaluation
 import numpy as np
 
+
 @unittest.skipIf(
     "CUVS" not in faiss.get_compile_options(),
     "only if cuVS is compiled in")
diff --git a/faiss/gpu/test/test_gpu_basics.py b/faiss/gpu/test/test_gpu_basics.py
@@ -257,6 +257,7 @@ def create_gpu(dim):
 
         index.train(training_data)
 
+
 def make_t(num, d, clamp=False, seed=None):
     rs = None
     if seed is None:
diff --git a/faiss/gpu/test/test_gpu_index_serialize.py b/faiss/gpu/test/test_gpu_index_serialize.py
@@ -7,6 +7,7 @@
 import numpy as np
 import faiss
 
+
 def make_t(num, d):
     rs = np.random.RandomState(123)
     return rs.rand(num, d).astype('float32')
diff --git a/faiss/gpu/test/torch_test_contrib_gpu.py b/faiss/gpu/test/torch_test_contrib_gpu.py
@@ -20,6 +20,7 @@ def to_column_major_torch(x):
         # was default setting before memory_format was introduced
         return x.t().clone().t()
 
+
 def to_column_major_numpy(x):
     return x.T.copy().T
 
diff --git a/faiss/python/class_wrappers.py b/faiss/python/class_wrappers.py
@@ -34,12 +34,14 @@
 # because it is unclear how the conversion should occur: with a view
 # (= cast) or conversion?
 
+
 def _check_dtype_uint8(codes):
     if codes.dtype != 'uint8':
         raise TypeError("Input argument %s must be ndarray of dtype "
                         " uint8, but found %s" % ("codes", codes.dtype))
     return np.ascontiguousarray(codes)
 
+
 def _numeric_to_str(numeric_type):
     if numeric_type == faiss.Float32:
         return 'float32'
@@ -68,6 +70,7 @@ def _np_dtype_to_faiss_type(dtype):
         raise ValueError("given numpy dtype should be either np.int64, np.float32, np.float16, np.int8, or np.uint8 to convert to faiss numeric types")
 
 
+
 def replace_method(the_class, name, replacement, ignore_missing=False):
     """ Replaces a method in a class with another version. The old method
     is renamed to method_name_c (because presumably it was implemented in C) """
@@ -1380,6 +1383,7 @@ def from_array(self, array, bias=None):
 # Syntatic sugar for QINCo and QINCoStep
 ######################################################
 
+
 def handle_QINCoStep(the_class):
     the_class.original_init = the_class.__init__
 
diff --git a/tests/test_binary_io.py b/tests/test_binary_io.py
@@ -13,6 +13,7 @@
 import os
 import tempfile
 
+
 def make_binary_dataset(d, nb, nt, nq):
     assert d % 8 == 0
     x = np.random.randint(256, size=(nb + nq + nt, int(d / 8))).astype('uint8')
diff --git a/tests/test_fast_scan_ivf.py b/tests/test_fast_scan_ivf.py
@@ -151,6 +151,7 @@ def verify_with_draws(testcase, Dref, Iref, Dnew, Inew):
             mask = Dref[i, :] == dis
             testcase.assertEqual(set(Iref[i, mask]), set(Inew[i, mask]))
 
+
 def three_metrics(Dref, Iref, Dnew, Inew):
     nq = Iref.shape[0]
     recall_at_1 = (Iref[:, 0] == Inew[:, 0]).sum() / nq
@@ -547,7 +548,7 @@ def test_by_residual_odd_dim(self):
 
 
 class TestReconstruct(unittest.TestCase):
-    """ test reconstruct and sa_encode / sa_decode 
+    """ test reconstruct and sa_encode / sa_decode
     (also for a few additive quantizer variants) """
 
     def do_test(self, by_residual=False):
@@ -576,7 +577,7 @@ def do_test(self, by_residual=False):
         index.reconstruct_orig_invlists()
         assert index.orig_invlists.compute_ntotal() == index.ntotal
 
-        # compare with non fast-scan index 
+        # compare with non fast-scan index
         index2 = faiss.IndexIVFPQ(
             index.quantizer, d, 50, d // 2, 4, metric)
         index2.by_residual = by_residual
diff --git a/tests/test_residual_quantizer.py b/tests/test_residual_quantizer.py
