Remove l2gv2.utils.progress module

Implement tqdm_close() to finish progressbars. The other functions
progress.reset() and progress.update() are not needed as we can
call tqdm directly.

docs/source/reference/l2gv2.rst

@@ -63,14 +63,6 @@ l2gv2.models module
    :undoc-members:
    :show-inheritance:
 
-l2gv2.progress module
----------------------
-
-.. automodule:: l2gv2.progress
-   :members:
-   :undoc-members:
-   :show-inheritance:
-
 l2gv2.sparsify module
 ---------------------
 

l2gv2/graphs/npgraph.py

@@ -26,9 +26,10 @@
 import numpy as np
 import torch
 import numba
+from tqdm import tqdm
 from numba.experimental import jitclass
 
-from ..utils import progress
+from .utils import tqdm_close
 from .graph import Graph
 
 
@@ -334,14 +335,14 @@ def _memmap_degree(edge_index, num_nodes):
     degree = np.zeros(num_nodes, dtype=np.int64)
     with numba.objmode:
         print("computing degrees")
-        pbar = progress.reset(edge_index.shape[1])
+        pbar = tqdm(total=edge_index.shape[1])
     for it, source in enumerate(edge_index[0]):
         degree[source] += 1
         if it % 1000000 == 0 and it > 0:
             with numba.objmode:
-                progress.update(pbar, 1000000)
+                pbar.update(1000000)
     with numba.objmode:
-        progress.close(pbar)
+        tqdm_close(pbar)
     return degree
 
 
@@ -458,7 +459,7 @@ def partition_graph_edges(self, partition, self_loops):
         num_edges = self.num_edges
         with numba.objmode:
             print("finding partition edges")
-            pbar = progress.reset(num_edges)
+            pbar = tqdm(total=num_edges)
         num_clusters = partition.max() + 1
         edge_counts = np.zeros((num_clusters, num_clusters), dtype=np.int64)
         for i, (source, target) in enumerate(self.edge_index.T):
@@ -468,9 +469,9 @@ def partition_graph_edges(self, partition, self_loops):
                 edge_counts[source, target] += 1
             if i % 1000000 == 0 and i > 0:
                 with numba.objmode:
-                    progress.update(pbar, 1000000)
+                    pbar.update(1000000)
         with numba.objmode:
-            progress.close(pbar)
+            tqdm_close(pbar)
         index = np.nonzero(edge_counts)
         partition_edges = np.vstack(index)
         weights = np.empty((len(index[0]),), dtype=np.int64)

l2gv2/patch/clustering.py

@@ -17,7 +17,6 @@
 import numba
 
 from local2global_embedding.network import TGraph, NPGraph
-from local2global_embedding import progress
 
 
 def distributed_clustering(graph: TGraph, beta, rounds=None, patience=3, min_samples=2) -> torch.Tensor:
@@ -136,7 +135,7 @@ def _fennel_clustering(edge_index, adj_index, num_nodes, num_clusters, load_limi
     deltas = - alpha * gamma * (partition_sizes ** (gamma - 1))
 
     with numba.objmode:
-        progress.reset(num_nodes)
+        pbar = tqdm(total=num_nodes)
 
     for it in range(num_iters):
         not_converged = 0
@@ -169,17 +168,17 @@ def _fennel_clustering(edge_index, adj_index, num_nodes, num_clusters, load_limi
             if i % 10000 == 0 and i > 0:
                 progress_it = i
                 with numba.objmode:
-                    progress.update(10000)
+                    pbar.update(10000)
         with numba.objmode:
-            progress.update(num_nodes - progress_it)
+            pbar.update(num_nodes - progress_it)
 
         print('iteration: ' + str(it) + ', not converged: ' + str(not_converged))
 
         if not_converged == 0:
             print(f'converged after {it} iterations.')
             break
     with numba.objmode:
-        progress.close()
+        tqdm_close(pbar)
 
     return clusters
 

l2gv2/utils.py

@@ -19,12 +19,17 @@
 #  SOFTWARE.
 """TODO: module docstring for utils.py"""
 
-from tempfile import TemporaryFile
 from time import perf_counter
+
+from tqdm import tqdm
 import torch
 import torch.nn
 
 
+def tqdm_close(t: tqdm):
+    t.update(t.total - t.n)
+    t.close()
+
 def speye(n: int, dtype: torch.dtype = torch.float) -> torch.Tensor:
     """identity matrix of dimension n as sparse_coo_tensor."""
     return torch.sparse_coo_tensor(