[singa] branch dev updated: add sparsification implementation for mnist

[zh...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

zhaojing pushed a commit to branch dev
in repository https://gitbox.apache.org/repos/asf/singa.git


The following commit(s) were added to refs/heads/dev by this push:
     new b59e1c53 add sparsification implementation for mnist
     new e066bb9e Merge pull request #1037 from lzjpaul/23-2-22-zj-dev
b59e1c53 is described below

commit b59e1c53139fb6f694c2e27f141d2ba5eaf43ee1
Author: zhaojing <zh...@comp.nus.edu.sg>
AuthorDate: Wed Feb 22 11:37:06 2023 +0800

    add sparsification implementation for mnist
---
 .../autograd/sparsification_mnist.py               | 45 ++++++++++++++++++++++
 1 file changed, 45 insertions(+)

diff --git a/examples/largedataset_cnn/autograd/sparsification_mnist.py b/examples/largedataset_cnn/autograd/sparsification_mnist.py
new file mode 100644
index 00000000..315605ac
--- /dev/null
+++ b/examples/largedataset_cnn/autograd/sparsification_mnist.py
@@ -0,0 +1,45 @@
+#
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+#
+
+from mnist_cnn import *
+import multiprocessing
+import sys
+
+if __name__ == '__main__':
+
+    # Generate a NCCL ID to be used for collective communication
+    nccl_id = singa.NcclIdHolder()
+
+    # Number of GPUs to be used
+    world_size = int(sys.argv[1])
+
+    # Use sparsification with parameters
+    topK = False  # When topK = False, Sparsification based on a constant absolute threshold
+    corr = True  # If True, uses local accumulate gradient for the correction
+    sparsThreshold = 0.05  # The constant absolute threshold for sparsification
+
+    process = []
+    for local_rank in range(0, world_size):
+        process.append(
+            multiprocessing.Process(target=train_mnist_cnn,
+                                    args=(True, local_rank, world_size, nccl_id,
+                                          sparsThreshold, topK, corr)))
+
+    for p in process:
+        p.start()