updated execute_or_node_remote() to only set score or predict results at an output node where applicable

Author: yuanchi2807

Diff for: codeflare/pipelines/Runtime.py

@@ -62,7 +62,7 @@ class ExecutionType(Enum):
 
 
 @ray.remote
-def execute_or_node_remote(node: dm.EstimatorNode, mode: ExecutionType, xy_ref: dm.XYRef):
+def execute_or_node_remote(node: dm.EstimatorNode, mode: ExecutionType, xy_ref: dm.XYRef, is_outputNode: bool):
     """
     Helper remote function that executes an OR node. As such, this is a remote task that runs the estimator
     in the provided mode with the data pointed to by XYRef. The key aspect to note here is the choice of input
@@ -107,9 +107,16 @@ def execute_or_node_remote(node: dm.EstimatorNode, mode: ExecutionType, xy_ref:
     elif mode == ExecutionType.SCORE:
         if base.is_classifier(estimator) or base.is_regressor(estimator):
             estimator = node.get_estimator()
-            score_ref = ray.put(estimator.score(X, y))
-            result = dm.XYRef(score_ref, score_ref, prev_node_ptr, prev_node_ptr, [xy_ref])
-            return result
+            if is_outputNode:
+                score_ref = ray.put(estimator.score(X, y))
+                result = dm.XYRef(score_ref, score_ref, prev_node_ptr, prev_node_ptr, [xy_ref])
+                return result
+            else:
+                res_xy = estimator.score(xy_list)
+                res_xref = ray.put(res_xy.get_x())
+                res_yref = ray.put(res_xy.get_y())
+                result = dm.XYRef(res_xref, res_yref, prev_node_ptr, prev_node_ptr, Xyref_list)
+                return result
         else:
             res_Xref = ray.put(estimator.transform(X))
             result = dm.XYRef(res_Xref, xy_ref.get_yref(), prev_node_ptr, prev_node_ptr, [xy_ref])
@@ -118,9 +125,17 @@ def execute_or_node_remote(node: dm.EstimatorNode, mode: ExecutionType, xy_ref:
     elif mode == ExecutionType.PREDICT:
         # Test mode does not clone as it is a simple predict or transform
         if base.is_classifier(estimator) or base.is_regressor(estimator):
-            predict_ref = ray.put(estimator.predict(X))
-            result = dm.XYRef(predict_ref, predict_ref, prev_node_ptr, prev_node_ptr, [xy_ref])
-            return result
+            if is_outputNode:
+                predict_ref = ray.put(estimator.predict(X))
+                result = dm.XYRef(predict_ref, predict_ref, prev_node_ptr, prev_node_ptr, [xy_ref])
+                return result
+            else:
+                res_xy = estimator.predict(xy_list)
+                res_xref = ray.put(res_xy.get_x())
+                res_yref = ray.put(res_xy.get_y())
+
+                result = dm.XYRef(res_xref, res_yref, prev_node_ptr, prev_node_ptr, Xyref_list)
+                return result
         else:
             res_Xref = ray.put(estimator.transform(X))
             result = dm.XYRef(res_Xref, xy_ref.get_yref(), prev_node_ptr, prev_node_ptr, [xy_ref])
@@ -147,7 +162,10 @@ def execute_or_node(node, pre_edges, edge_args, post_edges, mode: ExecutionType)
         exec_xyrefs = []
         for xy_ref_ptr in Xyref_ptrs:
             xy_ref = ray.get(xy_ref_ptr)
-            inner_result = execute_or_node_remote.remote(node, mode, xy_ref)
+            if post_edges:
+                inner_result = execute_or_node_remote.remote(node, mode, xy_ref, True)
+            else:
+                inner_result = execute_or_node_remote.remote(node, mode, xy_ref, False)
             exec_xyrefs.append(inner_result)
 
         for post_edge in post_edges:

Diff for: notebooks/plot_nca_classification.ipynb

@@ -150,14 +150,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [
     {
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "2021-07-19 10:50:40,647\tINFO services.py:1267 -- View the Ray dashboard at \u001b[1m\u001b[32mhttp://127.0.0.1:8266\u001b[39m\u001b[22m\n"
+      "2021-07-22 17:14:51,530\tINFO services.py:1267 -- View the Ray dashboard at \u001b[1m\u001b[32mhttp://127.0.0.1:8266\u001b[39m\u001b[22m\n"
      ]
     },
     {
@@ -295,6 +295,13 @@
     "\n",
     "ray.shutdown()"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {

Diff for: notebooks/plot_rbm_logistic_classification.ipynb

@@ -27,16 +27,16 @@
      "output_type": "stream",
      "text": [
       "Automatically created module for IPython interactive environment\n",
-      "[BernoulliRBM] Iteration 1, pseudo-likelihood = -25.57, time = 0.10s\n",
-      "[BernoulliRBM] Iteration 2, pseudo-likelihood = -23.68, time = 0.14s\n",
-      "[BernoulliRBM] Iteration 3, pseudo-likelihood = -22.74, time = 0.15s\n",
-      "[BernoulliRBM] Iteration 4, pseudo-likelihood = -21.83, time = 0.14s\n",
-      "[BernoulliRBM] Iteration 5, pseudo-likelihood = -21.62, time = 0.15s\n",
-      "[BernoulliRBM] Iteration 6, pseudo-likelihood = -21.11, time = 0.10s\n",
-      "[BernoulliRBM] Iteration 7, pseudo-likelihood = -20.88, time = 0.13s\n",
-      "[BernoulliRBM] Iteration 8, pseudo-likelihood = -20.58, time = 0.13s\n",
-      "[BernoulliRBM] Iteration 9, pseudo-likelihood = -20.32, time = 0.13s\n",
-      "[BernoulliRBM] Iteration 10, pseudo-likelihood = -20.13, time = 0.13s\n",
+      "[BernoulliRBM] Iteration 1, pseudo-likelihood = -25.57, time = 0.13s\n",
+      "[BernoulliRBM] Iteration 2, pseudo-likelihood = -23.68, time = 0.18s\n",
+      "[BernoulliRBM] Iteration 3, pseudo-likelihood = -22.74, time = 0.18s\n",
+      "[BernoulliRBM] Iteration 4, pseudo-likelihood = -21.83, time = 0.17s\n",
+      "[BernoulliRBM] Iteration 5, pseudo-likelihood = -21.62, time = 0.17s\n",
+      "[BernoulliRBM] Iteration 6, pseudo-likelihood = -21.11, time = 0.18s\n",
+      "[BernoulliRBM] Iteration 7, pseudo-likelihood = -20.88, time = 0.17s\n",
+      "[BernoulliRBM] Iteration 8, pseudo-likelihood = -20.58, time = 0.17s\n",
+      "[BernoulliRBM] Iteration 9, pseudo-likelihood = -20.32, time = 0.17s\n",
+      "[BernoulliRBM] Iteration 10, pseudo-likelihood = -20.13, time = 0.16s\n",
       "Logistic regression using RBM features:\n",
       "              precision    recall  f1-score   support\n",
       "\n",
@@ -214,23 +214,23 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "2021-07-19 11:01:26,551\tINFO services.py:1267 -- View the Ray dashboard at \u001b[1m\u001b[32mhttp://127.0.0.1:8266\u001b[39m\u001b[22m\n"
+      "2021-07-22 17:16:19,742\tINFO services.py:1267 -- View the Ray dashboard at \u001b[1m\u001b[32mhttp://127.0.0.1:8267\u001b[39m\u001b[22m\n"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 1, pseudo-likelihood = -25.57, time = 0.11s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 2, pseudo-likelihood = -23.68, time = 0.15s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 3, pseudo-likelihood = -22.74, time = 0.15s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 4, pseudo-likelihood = -21.83, time = 0.16s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 5, pseudo-likelihood = -21.62, time = 0.15s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 6, pseudo-likelihood = -21.11, time = 0.14s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 7, pseudo-likelihood = -20.88, time = 0.13s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 8, pseudo-likelihood = -20.58, time = 0.15s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 9, pseudo-likelihood = -20.32, time = 0.15s\n",
-      "\u001b[2m\u001b[36m(pid=8995)\u001b[0m [BernoulliRBM] Iteration 10, pseudo-likelihood = -20.13, time = 0.15s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 1, pseudo-likelihood = -25.57, time = 0.16s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 2, pseudo-likelihood = -23.68, time = 0.22s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 3, pseudo-likelihood = -22.74, time = 0.22s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 4, pseudo-likelihood = -21.83, time = 0.22s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 5, pseudo-likelihood = -21.62, time = 0.22s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 6, pseudo-likelihood = -21.11, time = 0.21s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 7, pseudo-likelihood = -20.88, time = 0.21s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 8, pseudo-likelihood = -20.58, time = 0.21s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 9, pseudo-likelihood = -20.32, time = 0.21s\n",
+      "\u001b[2m\u001b[36m(pid=12180)\u001b[0m [BernoulliRBM] Iteration 10, pseudo-likelihood = -20.13, time = 0.47s\n",
       "Logistic regression using RBM features:\n",
       "              precision    recall  f1-score   support\n",
       "\n",