Updated the Python interface after recent changes to TreePPL (#7)

* Updated the serialization and deserialization of TreePPL matrices.

* Updated TreePPL examples after recent changes to the TreePPL syntax.

Author: kudlicka

examples/coin.tppl

@@ -1,4 +1,4 @@
-model function coin(outcomes: Bool[]): Real {
+model function coin(outcomes: Bool[]) => Real {
   assume p ~ Uniform(0.0, 1.0);
   for i in 1 to (length(outcomes)) {
     observe outcomes[i] ~ Bernoulli(p);

examples/crbd.tppl

@@ -30,7 +30,7 @@ function walk(node: Tree, time:Real, lambda: Real, mu: Real) {
   }
 }
 
-model function crbd(tree: Tree): Real[] {
+model function crbd(tree: Tree) => Real[] {
   assume lambda ~ Gamma(1.0, 1.0);
   assume mu ~ Gamma(1.0, 0.5);
   walk(tree.left, tree.age, lambda, mu);

examples/generative_crbd.tppl

@@ -1,4 +1,4 @@
-model function generativeCrbd(time: Real, lambda: Real, mu: Real): Tree {
+model function generativeCrbd(time: Real, lambda: Real, mu: Real) => Tree {
   assume waitingTime ~ Exponential(lambda + mu);
   let eventTime = time - waitingTime;
   if eventTime < 0.0 {

examples/treeppl_in_jupyter.ipynb

@@ -73,7 +73,7 @@
    "source": [
     "%%treeppl flip samples=10\n",
     "\n",
-    "model function flip(): Bool {\n",
+    "model function flip() => Bool {\n",
     "  assume p ~ Bernoulli(0.5);\n",
     "  return p;\n",
     "}"
@@ -102,35 +102,6 @@
     "res.samples"
    ]
   },
-  {
-   "cell_type": "markdown",
-   "id": "dcd09ad5",
-   "metadata": {},
-   "source": [
-    "We can dynamically adjust the number of samples as follows:"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "b5810a7c",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "flip.set_samples(1000)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "d961ee99",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "res = flip()\n",
-    "sns.countplot(y=res.samples)"
-   ]
-  },
   {
    "cell_type": "markdown",
    "id": "1e839e5b",
@@ -156,7 +127,7 @@
    "source": [
     "%%treeppl coin samples=100000\n",
     "\n",
-    "model function coin(outcomes: Bool[]): Real {\n",
+    "model function coin(outcomes: Bool[]) => Real {\n",
     "  assume p ~ Uniform(0.0, 1.0);\n",
     "  for i in 1 to (length(outcomes)) {\n",
     "    observe outcomes[i] ~ Bernoulli(p);\n",
@@ -200,7 +171,7 @@
    "source": [
     "%%treeppl generative_crbd samples=1\n",
     "\n",
-    "model function generativeCrbd(time: Real, lambda: Real, mu: Real): Tree {\n",
+    "model function generativeCrbd(time: Real, lambda: Real, mu: Real) => Tree {\n",
     "  assume waitingTime ~ Exponential(lambda + mu);\n",
     "  let eventTime = time - waitingTime;\n",
     "  if eventTime < 0.0 {\n",
@@ -291,7 +262,7 @@
     "  }\n",
     "}\n",
     "\n",
-    "model function crbd(tree: Tree): Real[] {\n",
+    "model function crbd(tree: Tree) => Real[] {\n",
     "  assume lambda ~ Gamma(1.0, 1.0);\n",
     "  assume mu ~ Gamma(1.0, 0.5);\n",
     "  walk(tree.left, tree.age, lambda, mu);\n",

treeppl/serialization.py

@@ -23,8 +23,8 @@ def object_hook(dictionary):
         return constructor_to_class[constructor](**dictionary.get("__data__", {}))
     if "__float__" in dictionary:
         return float(dictionary["__float__"])
-    if "__tensor__" in dictionary:
-        return np.array(dictionary["__tensor__"]).reshape(dictionary["__tensorShape__"])
+    if set(dictionary.keys()) == {"m", "n", "arr"}:
+        return np.array(dictionary["arr"]).reshape(dictionary["m"], dictionary["n"])
     return dictionary
 
 
@@ -46,8 +46,9 @@ def default(self, obj):
         try:
             if isinstance(obj, np.ndarray):
                 return {
-                    "__tensor__": obj.flatten().tolist(),
-                    "__tensorShape__": obj.shape,
+                    "m": obj.shape[0],
+                    "n": obj.shape[1],
+                    "arr": obj.flatten().tolist(),
                 }
             return json.JSONEncoder.default(self, obj)
         except TypeError: