feat: first test in ray

Author: dbuchaca

python_libraries/ray/01_parallel_calls.py

@@ -0,0 +1,47 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import ray
+import time
+from time import sleep
+
+ray.init(num_cpus=10, ignore_reinit_error=True)
+print('\nSuccessfully imported ray!')
+
+
+def regular_function():
+    sleep(2)
+    return 1
+
+@ray.remote
+def remote_function():
+    sleep(2)
+    return 1
+
+def sum_serial(n):
+    result = 0
+    for _ in range(n):
+        result += regular_function()
+    return result
+
+
+def sum_remote(n):
+    results = []
+    for _ in range(n):
+        results.append(remote_function.remote())
+    return sum(ray.get(results))
+
+
+n = 10
+
+t0 = time.time()
+print(f"\nsum_serial(n)={sum_serial(n)}")
+total_time = time.time() - t0
+print(f"time sum_serial={total_time}")
+
+
+t0 = time.time()
+print(f"\nsum_remote(n)={sum_remote(n)}")
+total_time = time.time() - t0
+print(f"time sum_remote={total_time}")

python_libraries/ray/README.md

@@ -0,0 +1,11 @@
+
+
+# Introduction to Ray
+
+
+Ray introduces the `ray.remote` decorator which turns a python function into a "remote function".
+Remote functions are "futures", a remote function `f` can runned with `f.remote()`.
+This will trigger the execution of the function but not block the runtime of the program (as an async function).
+
+In `01_parallel_calls.py` one can see the execution of a remote function, where each call needs 2 seconds.
+The remote function runned 10 times takes approximately 2 seconds, whereas the serial takes around 20.