refactor qnlp-snake-removal

discopy · Feb 11, 2020 · 4fd416a · 4fd416a
1 parent abdf7cd
commit 4fd416a
Show file tree

Hide file tree

Showing 3 changed files with 102 additions and 97 deletions.
diff --git a/__init__.py b/__init__.py
@@ -30,4 +30,4 @@
 from discopy.circuit import Circuit, Gate, Bra, Ket, CircuitFunctor
 from discopy.pregroup import Word
 
-__version__ = '0.2.2'
+__version__ = '0.2.3'
diff --git a/notebooks/qnlp-experiment.ipynb b/notebooks/qnlp-experiment.ipynb
@@ -39,7 +39,7 @@
       "Alice who loves Alice is rich.\n",
       "Alice who loves Bob is rich.\n",
       "\n",
-      "9.83 seconds to generate 10 sentences.\n"
+      "8.18 seconds to generate 10 sentences.\n"
      ]
     }
    ],
@@ -235,7 +235,7 @@
        "            └───┘                              └───┘     </pre>"
       ],
       "text/plain": [
-       "<qiskit.visualization.text.TextDrawing at 0x134d04a90>"
+       "<qiskit.visualization.text.TextDrawing at 0x1280c8fd0>"
       ]
      },
      "execution_count": 6,
@@ -279,8 +279,8 @@
     "from pytket.backends.ibm import AerBackend\n",
     "from discopy.circuit import matrix_from_counts\n",
     "\n",
-    "def evaluate(F, sentences, backend=AerBackend(noise_model), n_shots=2**10, seed=42):\n",
-    "    circuits = [F(parsing[s]).to_tk().measure_all() for s in sentences]\n",
+    "def evaluate(params, sentences, backend=AerBackend(noise_model), n_shots=2**10, seed=42):\n",
+    "    circuits = [F(params)(parsing[s]).to_tk().measure_all() for s in sentences]\n",
     "    list(map(backend.default_compilation_pass.apply, circuits))\n",
     "    backend.process_circuits(circuits, n_shots=n_shots, seed=seed)\n",
     "    return [matrix_from_counts(\n",
@@ -299,7 +299,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "3.275 seconds to compute the corpus.\n",
+      "3.175 seconds to compute the corpus.\n",
       "\n",
       "True sentences:\n",
       "Alice is rich. (0.760)\n",
@@ -323,7 +323,7 @@
     "from time import time\n",
     "\n",
     "start = time()\n",
-    "corpus = dict(zip(sentences, evaluate(F(params0), sentences)))\n",
+    "corpus = dict(zip(sentences, evaluate(params0, sentences)))\n",
     "print(\"{:.3f} seconds to compute the corpus.\\n\".format(time() - start))\n",
     "\n",
     "delta = .1\n",
@@ -372,7 +372,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": 14,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -381,12 +381,12 @@
     "def loss(params, sentences=sentence_train):\n",
     "    return - np.mean(np.array([\n",
     "        (corpus[sentence] - scalar) ** 2\n",
-    "        for sentence, scalar in zip(sentences, evaluate(F(params), sentences))]))"
+    "        for sentence, scalar in zip(sentences, evaluate(params, sentences))]))"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 12,
+   "execution_count": 15,
    "metadata": {},
    "outputs": [
     {
@@ -409,29 +409,29 @@
     "\n",
     "print(\"\\nIs Alice who loves Bob rich?\")\n",
     "print(\"Yes, she is.\"\n",
-    "      if evaluate(F(params), ['Alice who loves Bob is rich.'])[0] > .5 + delta\n",
+    "      if evaluate(params, ['Alice who loves Bob is rich.'])[0] > .5 + delta\n",
     "      else \"No, she isn't.\")"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 18,
+   "execution_count": 16,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Epoch 1 (4 seconds since start): [0.62252593 0.90506185]\n",
-      "Epoch 2 (8 seconds since start): [0.59053632 0.93705145]\n",
-      "Epoch 3 (11 seconds since start): [0.55806232 0.96952546]\n",
-      "Epoch 4 (15 seconds since start): [0.54260212 0.98498566]\n",
-      "Epoch 5 (19 seconds since start): [0.53112536 0.99646241]\n",
-      "Epoch 6 (22 seconds since start): [0.5233175  1.00427028]\n",
-      "Epoch 7 (26 seconds since start): [0.52084204 1.00674574]\n",
-      "Epoch 8 (30 seconds since start): [0.52248008 1.0051077 ]\n",
-      "Epoch 9 (33 seconds since start): [0.52288718 1.0047006 ]\n",
-      "Epoch 10 (38 seconds since start): [0.52284507 1.00474271]\n"
+      "Epoch 1 (3 seconds since start): [0.62252593 0.90506185]\n",
+      "Epoch 2 (6 seconds since start): [0.63414504 0.91668096]\n",
+      "Epoch 3 (8 seconds since start): [0.5855492 0.9652768]\n",
+      "Epoch 4 (11 seconds since start): [0.56230937 0.98851663]\n",
+      "Epoch 5 (14 seconds since start): [0.54748535 1.00334065]\n",
+      "Epoch 6 (16 seconds since start): [0.54391572 1.00691028]\n",
+      "Epoch 7 (18 seconds since start): [0.53935373 1.01147227]\n",
+      "Epoch 8 (20 seconds since start): [0.54198379 1.0088422 ]\n",
+      "Epoch 9 (22 seconds since start): [0.53921175 1.01161424]\n",
+      "Epoch 10 (24 seconds since start): [0.52035009 0.99275258]\n"
      ]
     },
     {
@@ -442,10 +442,10 @@
        "    nfev: 20\n",
        "     nit: 10\n",
        " success: True\n",
-       "       x: array([0.52284507, 1.00474271])"
+       "       x: array([0.52035009, 0.99275258])"
       ]
      },
-     "execution_count": 18,
+     "execution_count": 16,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -467,14 +467,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 19,
+   "execution_count": 17,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Testing loss: -0.011108393780887127\n"
+      "Testing loss: -0.011193843558430672\n"
      ]
     }
    ],
@@ -484,7 +484,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [
     {
@@ -499,7 +499,7 @@
    "source": [
     "print(\"Is Alice who loves Bob rich?\")\n",
     "print(\"Yes, she is.\"\n",
-    "      if evaluate(F(result.x), ['Alice who loves Bob is rich.'])[0] > .5 + delta\n",
+    "      if evaluate(result.x, ['Alice who loves Bob is rich.'])[0] > .5 + delta\n",
     "      else \"No, she isn't.\")"
    ]
   }