Release 0.3.3 (#49)

microsoft · Jun 27, 2023 · 9fcb726 · 9fcb726
1 parent 8e9cc2d
commit 9fcb726
Show file tree

Hide file tree

Showing 38 changed files with 1,191 additions and 1,050 deletions.
diff --git a/examples/__init__.py b/examples/__init__.py
diff --git a/examples/csuite_example.ipynb b/examples/csuite_example.ipynb
diff --git a/examples/multi_investment_sales_attribution.ipynb b/examples/multi_investment_sales_attribution.ipynb
@@ -1,6 +1,7 @@
 {
  "cells": [
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -20,6 +21,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -55,21 +57,19 @@
     "from pytorch_lightning.callbacks import TQDMProgressBar\n",
     "from tensordict import TensorDict\n",
     "\n",
-    "from causica.distributions import (\n",
-    "    ContinuousNoiseDist,\n",
-    "    SEMDistributionModule,\n",
-    ")\n",
+    "from causica.distributions import ContinuousNoiseDist\n",
     "from causica.lightning.data_modules.basic_data_module import BasicDECIDataModule\n",
     "from causica.lightning.modules.deci_module import DECIModule\n",
-    "from causica.sem.distribution_parameters_sem import DistributionParametersSEM\n",
+    "from causica.sem.sem_distribution import SEMDistributionModule\n",
     "from causica.sem.structural_equation_model import ite\n",
     "from causica.training.auglag import AugLagLRConfig\n",
     "\n",
     "warnings.filterwarnings(\"ignore\")\n",
-    "%matplotlib inline"
+    "test_run = bool(os.environ.get(\"TEST_RUN\", False))  # used by testing to run the notebook as a script"
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -110,6 +110,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -384,6 +385,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -439,21 +441,27 @@
     "fig, axis = plt.subplots(1, 1, figsize=(8, 8))\n",
     "labels = {node: i for i, node in enumerate(true_adj.nodes)}\n",
     "\n",
-    "layout = nx.nx_agraph.graphviz_layout(true_adj, prog=\"dot\")\n",
+    "try:\n",
+    "    layout = nx.nx_agraph.graphviz_layout(true_adj, prog=\"dot\")\n",
+    "except (ModuleNotFoundError, ImportError):\n",
+    "    layout = nx.layout.spring_layout(true_adj)\n",
+    "\n",
     "for node, i in labels.items():\n",
     "    axis.scatter(layout[node][0], layout[node][1], label=f\"{i}: {node}\")\n",
     "axis.legend()\n",
     "nx.draw_networkx(true_adj, pos=layout, with_labels=True, arrows=True, labels=labels, ax=axis)"
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
     "# Discover the Causal Graph"
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -478,6 +486,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -497,6 +506,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -523,6 +533,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -542,6 +553,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -553,6 +565,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -596,7 +609,8 @@
     "\n",
     "trainer = pl.Trainer(\n",
     "    accelerator=\"auto\",\n",
-    "    max_epochs=int(os.environ.get(\"MAX_EPOCH\", 2000)),  # used by testing to run the notebook as a script\n",
+    "    max_epochs=2000,\n",
+    "    fast_dev_run=test_run,\n",
     "    callbacks=[TQDMProgressBar(refresh_rate=19)],\n",
     "    enable_checkpointing=False,\n",
     ")"
@@ -668,6 +682,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -725,6 +740,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -733,6 +749,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -742,6 +759,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -769,7 +787,7 @@
    ],
    "source": [
     "revenue_estimated_ate = {}\n",
-    "num_samples = 20000\n",
+    "num_samples = 10 if test_run else 20000\n",
     "sample_shape = torch.Size([num_samples])\n",
     "transform = data_module.normalizer.transform_modules[outcome]().inv\n",
     "\n",
@@ -791,6 +809,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -825,6 +844,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "metadata": {},
    "source": [
@@ -857,13 +877,14 @@
    "source": [
     "revenue_estimated_ite = {}\n",
     "\n",
+    "base_noise = sem.sample_to_noise(data_module.dataset_train)\n",
+    "\n",
     "for treatment in treatment_columns:\n",
-    "    base_noise = sem.sample_to_noise(data_module.dataset_train)\n",
-    "    intervention_a = TensorDict({treatment: torch.tensor([1.0])}, batch_size=tuple())\n",
-    "    do_a_cfs = transform(sem.do(interventions=intervention_a).noise_to_sample(base_noise)[outcome])\n",
-    "    intervention_b = TensorDict({treatment: torch.tensor([0.0])}, batch_size=tuple())\n",
-    "    do_b_cfs = transform(sem.do(interventions=intervention_b).noise_to_sample(base_noise)[outcome])\n",
-    "    revenue_estimated_ite[treatment] = (do_a_cfs - do_b_cfs).cpu().detach().numpy()[:, 0]\n",
+    "    do_sem = sem.do(interventions=TensorDict({treatment: torch.tensor([1.0])}, batch_size=tuple()))\n",
+    "    do_a_cfs = transform(do_sem.noise_to_sample(base_noise)[outcome]).cpu().detach().numpy()[:, 0]\n",
+    "    do_sem = sem.do(interventions=TensorDict({treatment: torch.tensor([0.0])}, batch_size=tuple()))\n",
+    "    do_b_cfs = transform(do_sem.noise_to_sample(base_noise)[outcome]).cpu().detach().numpy()[:, 0]\n",
+    "    revenue_estimated_ite[treatment] = do_a_cfs - do_b_cfs\n",
     "\n",
     "revenue_estimated_ite"
    ]