{"behavior":"degenerate_repetition","label_filter":null,"limit":10,"next_offset":30973,"offset":30963,"previous_offset":null,"rows":[{"assistant_initiates":false,"asst_text":"Of course! Here's a simpler version of the prayer:\n\n---\n\nDear Mother Mary,\n\nIn threatening storms, we seek your protection. Keep us safe from harm and guide us with hope and peace. Please intercede with your Son, Jesus, to calm the storms around and within us.\n\nAmen.\n\n---","behavior":"degenerate_repetition","conversation_hash":"e015e94a62fd06ac4688716b6e780351","conversation_redacted":false,"conversation_toxic":false,"country":"Philippines","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":285,"n_turns_in_conversation":12,"n_words":50,"source_row_offset":30963,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:02 GMT","turn_id":"a2df713e7cd72320a07048f0","turn_index":5,"user_text":"other simple"},{"assistant_initiates":false,"asst_text":"Certainly! Here's an even simpler version:\n\n---\n\nDear Mother Mary,\n\nProtect us during this storm. Keep us safe and bring us peace.\n\nAmen.","behavior":"degenerate_repetition","conversation_hash":"e015e94a62fd06ac4688716b6e780351","conversation_redacted":false,"conversation_toxic":false,"country":"Philippines","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":152,"n_turns_in_conversation":12,"n_words":26,"source_row_offset":30964,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:02 GMT","turn_id":"b58fa8de34d2773d89036955","turn_index":6,"user_text":"make it simple"},{"assistant_initiates":false,"asst_text":"Of course! Here is the prayer rephrased in simpler English:\n\n---\n\nDear Blessed Mother Mary,\n\nWhen storms are strong and waters rise, we ask for your protection. Please wrap us in your love and keep us safe from danger. Ask your Son, Jesus, to calm the storm and bring peace to our hearts. Help us to stay hopeful and strong through these tough times.\n\nAmen.","behavior":"degenerate_repetition","conversation_hash":"e015e94a62fd06ac4688716b6e780351","conversation_redacted":false,"conversation_toxic":false,"country":"Philippines","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":701,"n_turns_in_conversation":12,"n_words":125,"source_row_offset":30965,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:02 GMT","turn_id":"db83730e4dde645dc28795b7","turn_index":7,"user_text":"Dear Blessed Mother Mary,\n\nIn times of threatening storms and rising waters, we turn to you for protection. Cover us with your mantle of love and keep us safe from harm. Intercede with your Son, Jesus, to calm the tempests and bring peace to our hearts. Guide us through these challenges with faith and hope.\n\nAmen. rephrase use simple english"},{"assistant_initiates":false,"asst_text":"A simple title for this prayer could be:\n\n---\n\n\"Prayer for Protection During Storms\"\n\n---","behavior":"degenerate_repetition","conversation_hash":"e015e94a62fd06ac4688716b6e780351","conversation_redacted":false,"conversation_toxic":false,"country":"Philippines","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":105,"n_turns_in_conversation":12,"n_words":18,"source_row_offset":30966,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:02 GMT","turn_id":"92727ea947700ce0641f329a","turn_index":8,"user_text":"title of prayer"},{"assistant_initiates":false,"asst_text":"Certainly! Here's the title with \"Mama Mary\" included:\n\n---\n\n\"Prayer to Mama Mary for Protection During Storms\"\n\n---","behavior":"degenerate_repetition","conversation_hash":"e015e94a62fd06ac4688716b6e780351","conversation_redacted":false,"conversation_toxic":false,"country":"Philippines","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":130,"n_turns_in_conversation":12,"n_words":21,"source_row_offset":30967,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:02 GMT","turn_id":"db7619ebaf74cd91db4be86a","turn_index":9,"user_text":"add mama mary"},{"assistant_initiates":false,"asst_text":"Certainly! Here's a prayer to Mama Mary for protection from a severe typhoon:\n\n---\n\n**Prayer to Mama Mary for Protection from a Typhoon**\n\nDear Mama Mary,\n\nAs a severe typhoon approaches, we seek your loving protection. Please shield our homes and loved ones from harm. Intercede with your Son, Jesus, to protect our families and properties. Grant us strength, peace, and safety during this time.\n\nAmen.","behavior":"degenerate_repetition","conversation_hash":"e015e94a62fd06ac4688716b6e780351","conversation_redacted":false,"conversation_toxic":false,"country":"Philippines","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":481,"n_turns_in_conversation":12,"n_words":79,"source_row_offset":30968,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:02 GMT","turn_id":"7fb2c2e95866b5917c76c9ae","turn_index":10,"user_text":"make evade from severe typhoon that may hit our properties and family members"},{"assistant_initiates":false,"asst_text":"Certainly! Here is a prayer to Saint Clare of Assisi for good weather and protection from severe weather:\n\n---\n\n**Prayer to Saint Clare of Assisi for Protection from Bad Weather**\n\nDear Saint Clare,\n\nYou who trusted God through every storm, we ask for your intercession for good weather. Please protect us from severe storms and keep our homes and loved ones safe. Guide us with your faith and help us find peace and comfort in God's care.\n\nAmen.","behavior":"degenerate_repetition","conversation_hash":"e015e94a62fd06ac4688716b6e780351","conversation_redacted":false,"conversation_toxic":false,"country":"Philippines","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":532,"n_turns_in_conversation":12,"n_words":94,"source_row_offset":30969,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:02 GMT","turn_id":"259f3e49384755ca6fcd8224","turn_index":11,"user_text":"make for saint claire of assisi for good weather and evade us from severe bad weather"},{"assistant_initiates":false,"asst_text":"Trong ng\u00f4n ng\u1eef l\u1eadp tr\u00ecnh Python, vi\u1ec7c x\u00e1c \u0111\u1ecbnh ki\u1ec3u d\u1eef li\u1ec7u c\u1ee7a m\u1ed9t bi\u1ebfn l\u00e0 r\u1ea5t quan tr\u1ecdng \u0111\u1ec3 \u0111\u1ea3m b\u1ea3o ch\u01b0\u01a1ng tr\u00ecnh ho\u1ea1t \u0111\u1ed9ng ch\u00ednh x\u00e1c. Python cung c\u1ea5p nhi\u1ec1u h\u00e0m t\u00edch h\u1ee3p s\u1eb5n gi\u00fap l\u1eadp tr\u00ecnh vi\u00ean ki\u1ec3m tra v\u00e0 thao t\u00e1c v\u1edbi d\u1eef li\u1ec7u hi\u1ec7u qu\u1ea3. M\u1ed9t trong nh\u1eefng h\u00e0m h\u1eefu \u00edch nh\u1ea5t \u0111\u1ec3 nh\u1eadn bi\u1ebft ki\u1ec3u d\u1eef li\u1ec7u c\u1ee7a bi\u1ebfn l\u00e0 h\u00e0m `type()`.\n\nH\u00e0m `type()` c\u00f3 ch\u1ee9c n\u0103ng tr\u1ea3 v\u1ec1 ki\u1ec3u d\u1eef li\u1ec7u c\u1ee7a \u0111\u1ed1i t\u01b0\u1ee3ng \u0111\u01b0\u1ee3c truy\u1ec1n v\u00e0o n\u00f3. Khi b\u1ea1n s\u1eed d\u1ee5ng h\u00e0m n\u00e0y, Python s\u1ebd tr\u1ea3 v\u1ec1 m\u1ed9t \u0111\u1ed1i t\u01b0\u1ee3ng ki\u1ec3u, ch\u1eb3ng h\u1ea1n nh\u01b0 `int`, `str`, `list`, `dict`, v\u00e0 nhi\u1ec1u ki\u1ec3u d\u1eef li\u1ec7u kh\u00e1c t\u00f9y thu\u1ed9c v\u00e0o gi\u00e1 tr\u1ecb c\u1ee7a bi\u1ebfn. \u0110\u00e2y l\u00e0 c\u00e1ch th\u1ee9c c\u01a1 b\u1ea3n \u0111\u1ec3 ki\u1ec3m tra lo\u1ea1i d\u1eef li\u1ec7u m\u00e0 b\u1ea1n \u0111ang l\u00e0m vi\u1ec7c, gi\u00fap x\u00e1c \u0111\u1ecbnh c\u00e1ch x\u1eed l\u00fd bi\u1ebfn \u0111\u00f3 trong c\u00e1c t\u00e1c v\u1ee5 l\u1eadp tr\u00ecnh ti\u1ebfp theo.\n\nV\u00ed d\u1ee5, n\u1ebfu b\u1ea1n c\u00f3 m\u1ed9t bi\u1ebfn `a = 10`, v\u00e0 b\u1ea1n mu\u1ed1n bi\u1ebft li\u1ec7u `a` c\u00f3 ph\u1ea3i l\u00e0 m\u1ed9t s\u1ed1 nguy\u00ean hay kh\u00f4ng, b\u1ea1n c\u00f3 th\u1ec3 s\u1eed d\u1ee5ng c\u00e2u l\u1ec7nh `type(a)`. K\u1ebft qu\u1ea3 tr\u1ea3 v\u1ec1 s\u1ebd l\u00e0 ``, cho bi\u1ebft r\u1eb1ng bi\u1ebfn `a` thu\u1ed9c ki\u1ec3u d\u1eef li\u1ec7u nguy\u00ean. T\u01b0\u01a1ng t\u1ef1, n\u1ebfu bi\u1ebfn `b = \"Hello, World!\"`, th\u00ec `type(b)` s\u1ebd tr\u1ea3 v\u1ec1 ``, ch\u1ec9 ra r\u1eb1ng bi\u1ebfn `b` l\u00e0 m\u1ed9t chu\u1ed7i k\u00fd t\u1ef1.\n\nVi\u1ec7c hi\u1ec3u r\u00f5 v\u00e0 x\u00e1c \u0111\u1ecbnh \u0111\u00fang ki\u1ec3u d\u1eef li\u1ec7u c\u1ee7a c\u00e1c bi\u1ebfn kh\u00f4ng ch\u1ec9 gi\u00fap ng\u0103n ng\u1eeba l\u1ed7i m\u00e0 c\u00f2n t\u1ed1i \u01b0u h\u00f3a hi\u1ec7u su\u1ea5t c\u1ee7a m\u00e3 ngu\u1ed3n. H\u01a1n n\u1eefa, khi x\u1eed l\u00fd d\u1eef li\u1ec7u ph\u1ee9c t\u1ea1p, nh\u01b0 s\u1eed d\u1ee5ng c\u00e1c c\u1ea5u tr\u00fac d\u1eef li\u1ec7u d\u1ea1ng danh s\u00e1ch ho\u1eb7c t\u1eeb \u0111i\u1ec3n, vi\u1ec7c bi\u1ebft ch\u00ednh x\u00e1c ki\u1ec3u d\u1eef li\u1ec7u gi\u00fap b\u1ea1n c\u00f3 th\u1ec3 \u00e1p d\u1ee5ng c\u00e1c ph\u01b0\u01a1ng th\u1ee9c, h\u00e0m li\u00ean quan m\u1ed9t c\u00e1ch hi\u1ec7u qu\u1ea3.\n\nT\u00f3m l\u1ea1i, vi\u1ec7c x\u00e1c \u0111\u1ecbnh ki\u1ec3u d\u1eef li\u1ec7u trong Python l\u00e0 b\u01b0\u1edbc c\u01a1 b\u1ea3n nh\u01b0ng v\u00f4 c\u00f9ng c\u1ea7n thi\u1ebft, \u0111\u1eb7c bi\u1ec7t khi ph\u00e1t tri\u1ec3n c\u00e1c \u1ee9ng d\u1ee5ng l\u1edbn. K\u1ef9 n\u0103ng n\u00e0y kh\u00f4ng ch\u1ec9 gi\u00fap l\u1eadp tr\u00ecnh vi\u00ean vi\u1ebft m\u00e3 ch\u00ednh x\u00e1c m\u00e0 c\u00f2n gi\u00fap h\u1ecd t\u00ecm ra gi\u1ea3i ph\u00e1p th\u00edch h\u1ee3p cho t\u1eebng lo\u1ea1i d\u1eef li\u1ec7u trong qu\u00e1 tr\u00ecnh x\u1eed l\u00fd th\u00f4ng tin.","behavior":"degenerate_repetition","conversation_hash":"3eaaddda6c4f861e5754d2898bc8bec9","conversation_redacted":false,"conversation_toxic":false,"country":"United States","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":2230,"n_turns_in_conversation":1,"n_words":463,"source_row_offset":30970,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:08 GMT","turn_id":"70800c13e80bfddb88035dc8","turn_index":0,"user_text":"Write a educational passage in Vietnamese that could appear in a textbook, giving some paragraphs or pharse that related to this question and answer set:\nC\u00e2u h\u1ecfi: L\u1ec7nh n\u00e0o d\u00f9ng \u0111\u1ec3 nh\u1eadn bi\u1ebft ki\u1ec3u d\u1eef li\u1ec7u c\u1ee7a bi\u1ebfn trong python?\nA. type().\nB. int().\nC. size().\nD. abs().\n\u0110\u00e1p \u00e1n \u0111\u00fang: type().\nNotes that:\n1. Remember that the multichoice question above must not be included in the paragraphs or phrase\n2. You can not take an example same as the question\n3. Explain detail like in the actual textbook\n4. All instructions must be given in general form"},{"assistant_initiates":false,"asst_text":"\u597d\u7684\uff0c\u6211\u4f1a\u6839\u636e\u60a8\u7684\u8981\u6c42\u4fee\u6539\u4ee3\u7801\uff0c\u5c06\u8bad\u7ec3\u96c6\u7684\u4e00\u90e8\u5206\u5212\u5206\u4e3a\u9a8c\u8bc1\u96c6\uff0c\u5e76\u68c0\u67e5\u4ee3\u7801\u4e2d\u662f\u5426\u5b58\u5728\u6570\u636e\u6cc4\u9732\u3002\n\n\u9996\u5148\uff0c\u6211\u4eec\u9700\u8981\u5728\u8bad\u7ec3\u96c6\u4e2d\u8fdb\u4e00\u6b65\u5212\u5206\u51fa\u9a8c\u8bc1\u96c6\u3002\u7136\u540e\uff0c\u786e\u4fdd\u6570\u636e\u9884\u5904\u7406\u6b65\u9aa4\uff08\u5982\u6807\u51c6\u5316\u548c\u5f52\u4e00\u5316\uff09\u53ea\u5728\u8bad\u7ec3\u96c6\u4e0a\u62df\u5408\uff0c\u5e76\u5e94\u7528\u4e8e\u9a8c\u8bc1\u96c6\u548c\u6d4b\u8bd5\u96c6\u3002\n\n\u4ee5\u4e0b\u662f\u4fee\u6539\u540e\u7684\u4ee3\u7801\uff1a\n\n```python\nimport sys\nimport os\nimport pennylane as qml\nfrom pennylane import numpy as np\nimport tensorflow as tf\nfrom sklearn.datasets import load_iris\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.preprocessing import OneHotEncoder, StandardScaler, MinMaxScaler\n\nimport matplotlib\nmatplotlib.use('Agg') # \u4f7f\u7528\u975e\u4ea4\u4e92\u5f0f\u540e\u7aef\uff0c\u4ee5\u4fbf\u5728\u6ca1\u6709\u663e\u793a\u73af\u5883\u7684\u60c5\u51b5\u4e0b\u8fd0\u884c\nimport matplotlib.pyplot as plt\n\nfrom tensorflow.keras.layers import Dense\nfrom importance_sampling.training import ConstantTimeImportanceTraining\nfrom importance_sampling.samplers import ConstantTimeSampler\nfrom importance_sampling.datasets import InMemoryDataset\nfrom importance_sampling.reweighting import BiasedReweightingPolicy\nfrom tensorflow.keras.callbacks import EarlyStopping\n\n# \u91cd\u5b9a\u5411\u6807\u51c6\u8f93\u51fa\u4ee5\u6355\u83b7\u6253\u5370\u8f93\u51fa\nclass Tee(object):\n def __init__(self, *files):\n self.files = files\n def write(self, obj):\n for f in self.files:\n f.write(obj)\n f.flush()\n def flush(self):\n for f in self.files:\n f.flush()\n\noutput_file = open('output_modified.log', 'w', encoding='utf-8')\noriginal_stdout = sys.stdout\nsys.stdout = Tee(sys.stdout, output_file)\n\ndef save_plot(fig, filename):\n fig.savefig(filename)\n plt.close(fig)\n\ndef plot_history(history):\n # \u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u51c6\u786e\u7387\n fig, ax = plt.subplots(figsize=(10, 5))\n ax.plot(history.history['accuracy'], label='Train Accuracy')\n if 'val_accuracy' in history.history:\n ax.plot(history.history['val_accuracy'], label='Validation Accuracy')\n else:\n print(\"No validation accuracy found in history.\")\n ax.set_title('Model Accuracy')\n ax.set_ylabel('Accuracy')\n ax.set_xlabel('Epoch')\n ax.legend(loc='upper left')\n save_plot(fig, 'accuracy_plot.png')\n\n # \u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u635f\u5931\n fig, ax = plt.subplots(figsize=(10, 5))\n ax.plot(history.history['loss'], label='Train Loss')\n if 'val_loss' in history.history:\n ax.plot(history.history['val_loss'], label='Validation Loss')\n else:\n print(\"No validation loss found in history.\")\n ax.set_title('Model Loss')\n ax.set_ylabel('Loss')\n ax.set_xlabel('Epoch')\n ax.legend(loc='upper right')\n save_plot(fig, 'loss_plot.png')\n\ndef calculate_accuracy_fluctuation(history):\n if 'val_accuracy' in history.history:\n accuracies = history.history['val_accuracy']\n elif 'accuracy' in history.history:\n accuracies = history.history['accuracy']\n else:\n print(\"No accuracy metric found in history.\")\n return None, None\n mean_accuracy = np.mean(accuracies)\n fluctuation = np.sum(np.abs(accuracies - mean_accuracy))\n return mean_accuracy, fluctuation\n\n# \u5b9a\u4e49\u5177\u67094\u4e2a\u91cf\u5b50\u6bd4\u7279\u7684\u91cf\u5b50\u8bbe\u5907\ndev = qml.device('default.qubit', wires=4)\n\n@qml.qnode(dev, interface='tf', diff_method='backprop')\ndef quantum_neural_network(inputs, weights):\n num_qubits = 4\n\n # \u5c06\u8f93\u5165\u6570\u636e\u7f16\u7801\u5230\u91cf\u5b50\u6001\u4e2d\n for i in range(num_qubits):\n qml.RY(inputs[i], wires=i)\n\n # \u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u7684\u53c2\u6570\u5316\u7ebf\u8def\n n_layers = weights.shape[0]\n for layer in range(n_layers):\n # \u5bf9\u6bcf\u4e2a\u91cf\u5b50\u6bd4\u7279\u5e94\u7528\u65cb\u8f6c\u95e8\n for i in range(num_qubits):\n params = weights[layer, i]\n qml.Rot(params[0], params[1], params[2], wires=i)\n # \u5e94\u7528\u7ea0\u7f20\u95e8\uff0c\u5982 CNOT \u95e8\n for i in range(num_qubits - 1):\n qml.CNOT(wires=[i, i + 1])\n\n # \u5728\u6bcf\u4e2a\u91cf\u5b50\u6bd4\u7279\u4e0a\u6d4b\u91cf PauliZ \u671f\u671b\u503c\n return [qml.expval(qml.PauliZ(i)) for i in range(num_qubits)]\n\n# \u5b9a\u4e49\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u5c42\nclass QuantumNeuralNetworkLayer(tf.keras.layers.Layer):\n def __init__(self, n_qubits, n_layers):\n super(QuantumNeuralNetworkLayer, self).__init__()\n self.n_qubits = n_qubits\n self.n_layers = n_layers\n\n # Initialize trainable weights with dtype=tf.float64\n init = tf.random_uniform_initializer(minval=0, maxval=2 * np.pi)\n self.q_weights = self.add_weight(\n shape=(n_layers, n_qubits, 3),\n initializer=init,\n trainable=True,\n name='q_weights',\n dtype=tf.float64\n )\n\n def call(self, inputs):\n def circuit(inputs):\n inputs = tf.cast(inputs, dtype=tf.float64)\n result = quantum_neural_network(inputs, self.q_weights)\n return result\n\n outputs = tf.map_fn(circuit, inputs, dtype=tf.float64)\n return outputs\n\n# \u5b9a\u4e49\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\nclass QuantumNeuralNetwork(tf.keras.Model):\n def __init__(self, n_qubits, n_layers, num_classes):\n super(QuantumNeuralNetwork, self).__init__()\n self.n_qubits = n_qubits\n self.n_layers = n_layers\n self.num_classes = num_classes\n\n self.quantum_layer = QuantumNeuralNetworkLayer(n_qubits, n_layers)\n self.output_layer = Dense(num_classes, activation='softmax', dtype=tf.float64)\n\n def call(self, inputs):\n q_output = self.quantum_layer(inputs)\n return self.output_layer(q_output)\n\ndef is_quantum_model(model):\n for layer in model.layers:\n if isinstance(layer, QuantumNeuralNetworkLayer):\n return True\n elif hasattr(layer, 'layers'):\n if is_quantum_model(layer):\n return True\n return False\n\nif __name__ == \"__main__\":\n # \u52a0\u8f7d\u6570\u636e\n iris = load_iris()\n X = iris['data'].astype(np.float64) # \u786e\u4fdd\u6570\u636e\u7c7b\u578b\u4e3a float64\n y = iris['target'].reshape(-1, 1)\n\n # \u5bf9\u6807\u7b7e\u8fdb\u884c One-Hot \u7f16\u7801\n encoder = OneHotEncoder(sparse_output=False)\n y_encoded = encoder.fit_transform(y).astype(np.float64)\n num_classes = y_encoded.shape[1]\n\n # \u9996\u5148\u62c6\u5206\u4e3a\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\n x_full_train, x_test, y_full_train, y_test = train_test_split(\n X, y_encoded, test_size=0.2, random_state=116, stratify=y_encoded\n )\n\n # \u5728\u8bad\u7ec3\u96c6\u4e2d\u8fdb\u4e00\u6b65\u62c6\u5206\u51fa\u9a8c\u8bc1\u96c6\n x_train, x_val, y_train, y_val = train_test_split(\n x_full_train, y_full_train, test_size=0.25, random_state=42, stratify=y_full_train\n )\n # \u73b0\u5728\uff0cx_train \u662f\u603b\u6570\u636e\u768460%\uff0cx_val \u662f20%\uff0cx_test \u662f20%\n\n # \u6570\u636e\u6807\u51c6\u5316\u548c\u5f52\u4e00\u5316\uff08\u4ec5\u5728\u8bad\u7ec3\u96c6\u4e0a\u62df\u5408\uff09\n scaler = StandardScaler()\n X_train_standardized = scaler.fit_transform(x_train).astype(np.float64)\n X_val_standardized = scaler.transform(x_val).astype(np.float64)\n X_test_standardized = scaler.transform(x_test).astype(np.float64)\n\n min_max_scaler = MinMaxScaler(feature_range=(0, np.pi))\n X_train_mapped = min_max_scaler.fit_transform(X_train_standardized).astype(np.float64)\n X_val_mapped = min_max_scaler.transform(X_val_standardized).astype(np.float64)\n X_test_mapped = min_max_scaler.transform(X_test_standardized).astype(np.float64)\n\n # \u8d85\u53c2\u6570\n learning_rate = 0.02 # \u5b66\u4e60\u7387\n epochs = 100 # \u8bad\u7ec3\u8f6e\u6570\n batch_size = 16 # \u6279\u91cf\u5927\u5c0f\n n_qubits = 4 # 4 \u4e2a\u91cf\u5b50\u6bd4\u7279\n n_layers = 3 # \u5c42\u6570\n\n model = QuantumNeuralNetwork(\n n_qubits=n_qubits,\n n_layers=n_layers,\n num_classes=num_classes\n )\n\n if is_quantum_model(model):\n print(\"\u8be5\u6a21\u578b\u5305\u542b\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\uff08QNN\uff09\u5c42\u3002\")\n else:\n print(\"\u8be5\u6a21\u578b\u4e0d\u5305\u542b\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\uff08QNN\uff09\u5c42\u3002\")\n\n # \u7ed8\u5236\u91cf\u5b50\u7535\u8def\u56fe\n sample_inputs = np.ones(4, dtype=np.float64) * (np.pi / 2)\n sample_weights = np.zeros((n_layers, n_qubits, 3), dtype=np.float64)\n\n print(\"\\n\u91cf\u5b50\u7535\u8def\u56fe\uff1a\")\n print(qml.draw(quantum_neural_network)(sample_inputs, sample_weights))\n\n steps_per_epoch = len(X_train_mapped) // batch_size\n total_steps = epochs * steps_per_epoch\n\n # \u521b\u5efa\u6570\u636e\u96c6\uff08\u5305\u542b\u9a8c\u8bc1\u96c6\uff09\n dataset = InMemoryDataset(\n X_train_mapped,\n y_train,\n X_val_mapped,\n y_val,\n categorical=False\n )\n\n # \u521d\u59cb\u5316\u91cd\u52a0\u6743\u7b56\u7565\n reweighting_policy = BiasedReweightingPolicy(k=1.0)\n\n # \u521d\u59cb\u5316\u91cd\u8981\u6027\u91c7\u6837\u8bad\u7ec3\n wrapped_model = ConstantTimeImportanceTraining(model, score=\"loss\")\n\n # \u4fee\u6539\u91c7\u6837\u5668\u4ee5\u4f7f\u7528\u52a8\u6001\u7684 alpha \u548c beta\n def custom_sampler(dataset, batch_size, steps_per_epoch, epochs):\n return ConstantTimeSampler(\n dataset,\n reweighting_policy,\n model=wrapped_model.model,\n B=64,\n alpha=0.6,\n beta_start=0.4,\n beta_frames=total_steps\n )\n\n # \u5c06\u81ea\u5b9a\u4e49\u91c7\u6837\u5668\u8d4b\u7ed9 wrapped_model\n wrapped_model.sampler = custom_sampler\n\n # \u4f7f\u7528 per-sample \u7684\u635f\u5931\u51fd\u6570\n loss_fn = tf.keras.losses.CategoricalCrossentropy(from_logits=False, reduction=tf.keras.losses.Reduction.NONE)\n\n # \u7f16\u8bd1\u6a21\u578b\n model.compile(\n optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),\n loss=loss_fn,\n metrics=['accuracy'],\n run_eagerly=True # \u5e2e\u52a9\u8c03\u8bd5\n )\n\n # \u521b\u5efa\u4fdd\u5b58\u8bad\u7ec3\u6307\u6807\u7684\u56de\u8c03\u7c7b\n class MetricsHistory(tf.keras.callbacks.Callback):\n def on_train_begin(self, logs=None):\n self.history = {'loss': [], 'accuracy': [], 'val_loss': [], 'val_accuracy': []}\n\n def on_epoch_end(self, epoch, logs=None):\n self.history['loss'].append(logs.get('loss'))\n self.history['accuracy'].append(logs.get('accuracy'))\n self.history['val_loss'].append(logs.get('val_loss'))\n self.history['val_accuracy'].append(logs.get('val_accuracy'))\n\n # \u5b9e\u4f8b\u5316\u56de\u8c03\u51fd\u6570\n metrics_history = MetricsHistory()\n\n # \u4f7f\u7528 fit_dataset \u8bad\u7ec3\u6a21\u578b\n history = wrapped_model.fit_dataset(\n dataset=dataset,\n batch_size=batch_size,\n epochs=epochs,\n steps_per_epoch=steps_per_epoch,\n verbose=1,\n callbacks=[metrics_history] # \u4f7f\u7528\u81ea\u5b9a\u4e49\u56de\u8c03\u51fd\u6570\n )\n\n # \u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\u6a21\u578b\n score = model.evaluate(X_test_mapped, y_test, verbose=0)\n print('\u6d4b\u8bd5\u635f\u5931:', score[0])\n print('\u6d4b\u8bd5\u51c6\u786e\u7387:', score[1])\n\n # \u4f7f\u7528\u81ea\u5b9a\u4e49\u56de\u8c03\u51fd\u6570\u4e2d\u7684\u5386\u53f2\u8bb0\u5f55\u7ed8\u5236\u56fe\u8868\n plot_history(metrics_history)\n\n mean_accuracy, fluctuation = calculate_accuracy_fluctuation(metrics_history)\n if mean_accuracy is not None and fluctuation is not None:\n print('\u9a8c\u8bc1\u96c6\u5e73\u5747\u51c6\u786e\u7387:', mean_accuracy)\n print('\u51c6\u786e\u7387\u6ce2\u52a8:', fluctuation)\n\n # \u5173\u95ed\u8f93\u51fa\u6587\u4ef6\u5e76\u6062\u590d\u6807\u51c6\u8f93\u51fa\n sys.stdout = original_stdout\n output_file.close()\n\n print(\"\u811a\u672c\u6267\u884c\u5b8c\u6bd5\u3002\u8bf7\u68c0\u67e5 'output_modified.log' \u4ee5\u83b7\u53d6\u6253\u5370\u8f93\u51fa\uff0c\u5e76\u5728\u5f53\u524d\u76ee\u5f55\u4e0b\u67e5\u770b\u4fdd\u5b58\u7684\u56fe\u8868\u3002\")\n```\n\n\u5728\u4e0a\u8ff0\u4ee3\u7801\u4e2d\uff0c\u6211\u8fdb\u884c\u4e86\u4ee5\u4e0b\u4fee\u6539\uff1a\n\n1. **\u5212\u5206\u9a8c\u8bc1\u96c6**\uff1a\u5728\u521d\u59cb\u8bad\u7ec3\u96c6\u4e0a\uff0c\u4f7f\u7528 `train_test_split` \u5c06\u4e00\u90e8\u5206\u6570\u636e\u5212\u5206\u4e3a\u9a8c\u8bc1\u96c6\u3002\n\n ```python\n # \u9996\u5148\u62c6\u5206\u4e3a\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\n x_full_train, x_test, y_full_train, y_test = train_test_split(\n X, y_encoded, test_size=0.2, random_state=116, stratify=y_encoded\n )\n\n # \u5728\u8bad\u7ec3\u96c6\u4e2d\u8fdb\u4e00\u6b65\u62c6\u5206\u51fa\u9a8c\u8bc1\u96c6\n x_train, x_val, y_train, y_val = train_test_split(\n x_full_train, y_full_train, test_size=0.25, random_state=42, stratify=y_full_train\n )\n ```\n\n2. **\u6570\u636e\u9884\u5904\u7406**\uff1a\u786e\u4fdd\u6807\u51c6\u5316\u548c\u5f52\u4e00\u5316\u53ea\u5728\u8bad\u7ec3\u96c6\u4e0a\u62df\u5408\uff0c\u5e76\u5e94\u7528\u4e8e\u9a8c\u8bc1\u96c6\u548c\u6d4b\u8bd5\u96c6\uff0c\u4ee5\u9632\u6b62\u6570\u636e\u6cc4\u9732\u3002\n\n ```python\n # \u6570\u636e\u6807\u51c6\u5316\u548c\u5f52\u4e00\u5316\uff08\u4ec5\u5728\u8bad\u7ec3\u96c6\u4e0a\u62df\u5408\uff09\n scaler = StandardScaler()\n X_train_standardized = scaler.fit_transform(x_train).astype(np.float64)\n X_val_standardized = scaler.transform(x_val).astype(np.float64)\n X_test_standardized = scaler.transform(x_test).astype(np.float64)\n\n min_max_scaler = MinMaxScaler(feature_range=(0, np.pi))\n X_train_mapped = min_max_scaler.fit_transform(X_train_standardized).astype(np.float64)\n X_val_mapped = min_max_scaler.transform(X_val_standardized).astype(np.float64)\n X_test_mapped = min_max_scaler.transform(X_test_standardized).astype(np.float64)\n ```\n\n3. **\u521b\u5efa\u6570\u636e\u96c6**\uff1a\u5728\u521b\u5efa `InMemoryDataset` \u65f6\uff0c\u5305\u542b\u9a8c\u8bc1\u96c6\u6570\u636e\u3002\n\n ```python\n # \u521b\u5efa\u6570\u636e\u96c6\uff08\u5305\u542b\u9a8c\u8bc1\u96c6\uff09\n dataset = InMemoryDataset(\n X_train_mapped,\n y_train,\n X_val_mapped,\n y_val,\n categorical=False\n )\n ```\n\n4. **\u8bad\u7ec3\u6a21\u578b**\uff1a\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u6a21\u578b\u5c06\u4f7f\u7528\u9a8c\u8bc1\u96c6\u6765\u8bc4\u4f30\u6027\u80fd\uff0c\u5e76\u5728 `MetricsHistory` \u56de\u8c03\u4e2d\u8bb0\u5f55\u8bad\u7ec3\u548c\u9a8c\u8bc1\u7684\u635f\u5931\u548c\u51c6\u786e\u7387\u3002\n\n ```python\n # \u521b\u5efa\u4fdd\u5b58\u8bad\u7ec3\u6307\u6807\u7684\u56de\u8c03\u7c7b\n class MetricsHistory(tf.keras.callbacks.Callback):\n def on_train_begin(self, logs=None):\n self.history = {'loss': [], 'accuracy': [], 'val_loss': [], 'val_accuracy': []}\n\n def on_epoch_end(self, epoch, logs=None):\n self.history['loss'].append(logs.get('loss'))\n self.history['accuracy'].append(logs.get('accuracy'))\n self.history['val_loss'].append(logs.get('val_loss'))\n self.history['val_accuracy'].append(logs.get('val_accuracy'))\n ```\n\n5. **\u7ed8\u5236\u56fe\u8868**\uff1a\u4fee\u6539\u7ed8\u56fe\u51fd\u6570\uff0c\u4ee5\u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u7684\u6307\u6807\u3002\n\n ```python\n def plot_history(history):\n # \u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u51c6\u786e\u7387\n fig, ax = plt.subplots(figsize=(10, 5))\n ax.plot(history.history['accuracy'], label='Train Accuracy')\n if 'val_accuracy' in history.history:\n ax.plot(history.history['val_accuracy'], label='Validation Accuracy')\n else:\n print(\"No validation accuracy found in history.\")\n ax.set_title('Model Accuracy')\n ax.set_ylabel('Accuracy')\n ax.set_xlabel('Epoch')\n ax.legend(loc='upper left')\n save_plot(fig, 'accuracy_plot.png')\n\n # \u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u635f\u5931\n fig, ax = plt.subplots(figsize=(10, 5))\n ax.plot(history.history['loss'], label='Train Loss')\n if 'val_loss' in history.history:\n ax.plot(history.history['val_loss'], label='Validation Loss')\n else:\n print(\"No validation loss found in history.\")\n ax.set_title('Model Loss')\n ax.set_ylabel('Loss')\n ax.set_xlabel('Epoch')\n ax.legend(loc='upper right')\n save_plot(fig, 'loss_plot.png')\n ```\n\n6. **\u68c0\u67e5\u6570\u636e\u6cc4\u9732**\uff1a\u901a\u8fc7\u786e\u4fdd\u6240\u6709\u7684\u6570\u636e\u9884\u5904\u7406\u6b65\u9aa4\u53ea\u5728\u8bad\u7ec3\u96c6\u4e0a\u62df\u5408\uff0c\u5e76\u5e94\u7528\u4e8e\u9a8c\u8bc1\u96c6\u548c\u6d4b\u8bd5\u96c6\uff0c\u6211\u4eec\u907f\u514d\u4e86\u6570\u636e\u6cc4\u9732\u3002\u6b64\u5916\uff0c\u6211\u4eec\u5728\u5212\u5206\u6570\u636e\u96c6\u65f6\uff0c\u4fdd\u6301\u4e86\u6570\u636e\u7684\u72ec\u7acb\u6027\uff0c\u6ca1\u6709\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u4f7f\u7528\u6d4b\u8bd5\u96c6\u7684\u6570\u636e\u3002\n\n7. **\u9a8c\u8bc1\u6a21\u578b\u662f\u5426\u4f7f\u7528\u4e86\u9a8c\u8bc1\u96c6**\uff1a\u901a\u8fc7\u5728 `fit_dataset` \u51fd\u6570\u4e2d\u5305\u542b\u9a8c\u8bc1\u96c6\u6570\u636e\uff0c\u5e76\u5728\u56de\u8c03\u51fd\u6570\u4e2d\u8bb0\u5f55\u9a8c\u8bc1\u96c6\u7684\u6307\u6807\uff0c\u6211\u4eec\u53ef\u4ee5\u786e\u4fdd\u6a21\u578b\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u6b63\u786e\u8bc4\u4f30\u4e86\u9a8c\u8bc1\u96c6\u7684\u6027\u80fd\u3002\n\n\u5e0c\u671b\u8fd9\u4e9b\u4fee\u6539\u6ee1\u8db3\u60a8\u7684\u8981\u6c42\u3002","behavior":"degenerate_repetition","conversation_hash":"e5b68f3a17dbcba366cc667a7c3d0e4e","conversation_redacted":false,"conversation_toxic":false,"country":"Singapore","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"o1-preview-2024-09-12","n_chars":30880,"n_turns_in_conversation":1,"n_words":2495,"source_row_offset":30971,"source_shard":"turns-00051.parquet","timestamp":"Thu, 14 Nov 2024 06:23:11 GMT","turn_id":"1f423dc0873367ec0d8e0196","turn_index":0,"user_text":"#\n# #\u4e0d\u4f7f\u7528\u9a8c\u8bc1\u96c6\n# import sys\n# import os\n# import pennylane as qml\n# from pennylane import numpy as np\n# import tensorflow as tf\n# from sklearn.datasets import load_iris\n# from sklearn.model_selection import train_test_split\n# from sklearn.preprocessing import OneHotEncoder, StandardScaler, MinMaxScaler\n# import matplotlib\n# matplotlib.use('Agg')\n# import matplotlib.pyplot as plt\n# from tensorflow.keras.layers import Dense\n# from importance_sampling.training import ConstantTimeImportanceTraining\n# from importance_sampling.samplers import ConstantTimeSampler\n# from importance_sampling.datasets import InMemoryDataset\n# from importance_sampling.reweighting import BiasedReweightingPolicy\n# from tensorflow.keras.callbacks import EarlyStopping\n#\n# # \u91cd\u5b9a\u5411\u6807\u51c6\u8f93\u51fa\u4ee5\u6355\u83b7\u6253\u5370\u8f93\u51fa\n# class Tee(object):\n# def __init__(self, *files):\n# self.files = files\n# def write(self, obj):\n# for f in self.files:\n# f.write(obj)\n# f.flush()\n# def flush(self):\n# for f in self.files:\n# f.flush()\n#\n# output_file = open('output_modified.log', 'w', encoding='utf-8')\n# original_stdout = sys.stdout\n# sys.stdout = Tee(sys.stdout, output_file)\n#\n# def save_plot(fig, filename):\n# fig.savefig(filename)\n# plt.close(fig)\n#\n# def plot_history(history):\n# # \u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u51c6\u786e\u7387\n# fig, ax = plt.subplots(figsize=(10, 5))\n# ax.plot(history.history['accuracy'], label='Train Accuracy')\n# # \u5982\u679c\u5b58\u5728\u9a8c\u8bc1\u51c6\u786e\u7387\uff0c\u7ed8\u5236\u9a8c\u8bc1\u51c6\u786e\u7387\n# if 'val_accuracy' in history.history:\n# ax.plot(history.history['val_accuracy'], label='Validation Accuracy')\n# ax.set_title('Model Accuracy')\n# ax.set_ylabel('Accuracy')\n# ax.set_xlabel('Epoch')\n# ax.legend(loc='upper left')\n# save_plot(fig, 'accuracy_plot.png')\n#\n# # \u7ed8\u5236\u8bad\u7ec3\u548c\u9a8c\u8bc1\u635f\u5931\n# fig, ax = plt.subplots(figsize=(10, 5))\n# ax.plot(history.history['loss'], label='Train Loss')\n# if 'val_loss' in history.history:\n# ax.plot(history.history['val_loss'], label='Validation Loss')\n# ax.set_title('Model Loss')\n# ax.set_ylabel('Loss')\n# ax.set_xlabel('Epoch')\n# ax.legend(loc='upper left')\n# save_plot(fig, 'loss_plot.png')\n#\n# def calculate_accuracy_fluctuation(history):\n# if 'val_accuracy' in history.history:\n# val_accuracies = history.history['val_accuracy']\n# else:\n# val_accuracies = history.history['accuracy']\n# mean_accuracy = np.mean(val_accuracies)\n# fluctuation = np.sum(np.abs(val_accuracies - mean_accuracy))\n# return mean_accuracy, fluctuation\n#\n# # \u5b9a\u4e49\u5177\u67094\u4e2a\u91cf\u5b50\u6bd4\u7279\u7684\u91cf\u5b50\u8bbe\u5907\n# dev = qml.device('default.qubit', wires=4)\n#\n# @qml.qnode(dev, interface='tf', diff_method='backprop')\n# def quantum_neural_network(inputs, weights):\n# num_qubits = 4\n#\n# # \u5c06\u8f93\u5165\u6570\u636e\u7f16\u7801\u5230\u91cf\u5b50\u6001\u4e2d\n# for i in range(num_qubits):\n# qml.RY(inputs[i], wires=i)\n#\n# # \u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u7684\u53c2\u6570\u5316\u7ebf\u8def\n# n_layers = weights.shape[0]\n# for layer in range(n_layers):\n# # \u5bf9\u6bcf\u4e2a\u91cf\u5b50\u6bd4\u7279\u5e94\u7528\u65cb\u8f6c\u95e8\n# for i in range(num_qubits):\n# params = weights[layer, i]\n# qml.Rot(params[0], params[1], params[2], wires=i)\n# # \u5e94\u7528\u7ea0\u7f20\u95e8\uff0c\u5982 CNOT \u95e8\n# for i in range(num_qubits - 1):\n# qml.CNOT(wires=[i, i + 1])\n#\n# # \u5728\u6bcf\u4e2a\u91cf\u5b50\u6bd4\u7279\u4e0a\u6d4b\u91cf PauliZ \u671f\u671b\u503c\n# return [qml.expval(qml.PauliZ(i)) for i in range(num_qubits)]\n#\n# # \u5b9a\u4e49\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u5c42\n# class QuantumNeuralNetworkLayer(tf.keras.layers.Layer):\n# def __init__(self, n_qubits, n_layers):\n# super(QuantumNeuralNetworkLayer, self).__init__()\n# self.n_qubits = n_qubits\n# self.n_layers = n_layers\n#\n# # Initialize trainable weights with dtype=tf.float64\n# init = tf.random_uniform_initializer(minval=0, maxval=2 * np.pi)\n# self.q_weights = self.add_weight(\n# shape=(n_layers, n_qubits, 3),\n# initializer=init,\n# trainable=True,\n# name='q_weights',\n# dtype=tf.float64\n# )\n#\n# def call(self, inputs):\n# def circuit(inputs):\n# inputs = tf.cast(inputs, dtype=tf.float64)\n# result = quantum_neural_network(inputs, self.q_weights)\n# return result\n#\n# outputs = tf.map_fn(circuit, inputs, dtype=tf.float64)\n# return outputs\n#\n# # \u5b9a\u4e49\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\n# class QuantumNeuralNetwork(tf.keras.Model):\n# def __init__(self, n_qubits, n_layers, num_classes):\n# super(QuantumNeuralNetwork, self).__init__()\n# self.n_qubits = n_qubits\n# self.n_layers = n_layers\n# self.num_classes = num_classes\n#\n# self.quantum_layer = QuantumNeuralNetworkLayer(n_qubits, n_layers)\n# self.output_layer = Dense(num_classes, activation='softmax', dtype=tf.float64)\n#\n# def call(self, inputs):\n# q_output = self.quantum_layer(inputs)\n# return self.output_layer(q_output)\n#\n# def is_quantum_model(model):\n# for layer in model.layers:\n# if isinstance(layer, QuantumNeuralNetworkLayer):\n# return True\n# elif hasattr(layer, 'layers'):\n# if is_quantum_model(layer):\n# return True\n# return False\n#\n# if __name__ == \"__main__\":\n# # \u52a0\u8f7d\u6570\u636e\n# iris = load_iris()\n# X = iris['data'].astype(np.float64) # \u786e\u4fdd\u6570\u636e\u7c7b\u578b\u4e3a float64\n# y = iris['target'].reshape(-1, 1)\n#\n# # \u5bf9\u6807\u7b7e\u8fdb\u884c One-Hot \u7f16\u7801\n# encoder = OneHotEncoder(sparse_output=False)\n# y_encoded = encoder.fit_transform(y).astype(np.float64)\n# num_classes = y_encoded.shape[1]\n#\n# # \u76f4\u63a5\u62c6\u5206\u4e3a\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\n# x_train, x_test, y_train, y_test = train_test_split(\n# X, y_encoded, test_size=0.2, random_state=116, stratify=y_encoded\n# )\n#\n# # \u6570\u636e\u6807\u51c6\u5316\u548c\u5f52\u4e00\u5316\n# scaler = StandardScaler()\n# X_train_standardized = scaler.fit_transform(x_train).astype(np.float64)\n# X_test_standardized = scaler.transform(x_test).astype(np.float64)\n#\n# min_max_scaler = MinMaxScaler(feature_range=(0, np.pi))\n# X_train_mapped = min_max_scaler.fit_transform(X_train_standardized).astype(np.float64)\n# X_test_mapped = min_max_scaler.transform(X_test_standardized).astype(np.float64)\n#\n# # \u8d85\u53c2\u6570\n# learning_rate = 0.05 # \u5b66\u4e60\u7387\n# epochs = 100 # \u589e\u52a0\u8bad\u7ec3\u8f6e\u6570\n# batch_size = 16 # \u6279\u91cf\u5927\u5c0f\n# n_qubits = 4 # 4 \u4e2a\u91cf\u5b50\u6bd4\u7279\n# n_layers = 3 # \u5c42\u6570\n#\n# model = QuantumNeuralNetwork(\n# n_qubits=n_qubits,\n# n_layers=n_layers,\n# num_classes=num_classes\n# )\n#\n# if is_quantum_model(model):\n# print(\"\u8be5\u6a21\u578b\u5305\u542b\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\uff08QNN\uff09\u5c42\u3002\")\n# else:\n# print(\"\u8be5\u6a21\u578b\u4e0d\u5305\u542b\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\uff08QNN\uff09\u5c42\u3002\")\n#\n# # \u7ed8\u5236\u91cf\u5b50\u7535\u8def\u56fe\n# sample_inputs = np.ones(4, dtype=np.float64) * (np.pi / 2)\n# sample_weights = np.zeros((n_layers, n_qubits, 3), dtype=np.float64)\n#\n# print(\"\\n\u91cf\u5b50\u7535\u8def\u56fe\uff1a\")\n# print(qml.draw(quantum_neural_network)(sample_inputs, sample_weights))\n#\n#\n# early_stopping = EarlyStopping(monitor='loss', patience=20, restore_best_weights=True)\n#\n# steps_per_epoch = len(X_train_mapped) // batch_size\n# total_steps = epochs * steps_per_epoch\n#\n# # \u4f7f\u7528\u91cd\u8981\u6027\u91c7\u6837\u8bad\u7ec3\n#\n# # \u521b\u5efa\u6570\u636e\u96c6\uff08\u65e0\u9a8c\u8bc1\u96c6\uff09\n# dataset = InMemoryDataset(\n# X_train_mapped,\n# y_train,\n# None, # \u65e0\u9a8c\u8bc1\u96c6\u6570\u636e\n# None,\n# categorical=False\n# )\n#\n# # \u521d\u59cb\u5316\u91cd\u52a0\u6743\u7b56\u7565\n# reweighting_policy = BiasedReweightingPolicy(k=1.0)\n#\n# # \u521d\u59cb\u5316\u91cd\u8981\u6027\u91c7\u6837\u8bad\u7ec3\n# wrapped_model = ConstantTimeImportanceTraining(model, score=\"loss\")\n#\n# # \u4fee\u6539\u91c7\u6837\u5668\u4ee5\u4f7f\u7528\u52a8\u6001\u7684 alpha \u548c beta\n# def custom_sampler(dataset, batch_size, steps_per_epoch, epochs):\n# return ConstantTimeSampler(\n# dataset,\n# reweighting_policy,\n# model=wrapped_model.model,\n# B=64,\n# alpha=0.6,\n# beta_start=0.4,\n# beta_frames=total_steps\n# )\n#\n# # \u5c06\u81ea\u5b9a\u4e49\u91c7\u6837\u5668\u8d4b\u7ed9 wrapped_model\n# wrapped_model.sampler = custom_sampler\n#\n# # \u4f7f\u7528 per-sample \u7684\u635f\u5931\u51fd\u6570\n# loss_fn = tf.keras.losses.CategoricalCrossentropy(from_logits=False, reduction=tf.keras.losses.Reduction.NONE)\n#\n# # \u7f16\u8bd1\u6a21\u578b\n# model.compile(\n# optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),\n# loss=loss_fn,\n# metrics=['accuracy'],\n# run_eagerly=True\n# )\n#\n# # \u4f7f\u7528 fit_dataset \u8bad\u7ec3\u6a21\u578b\n# history = wrapped_model.fit_dataset(\n# dataset=dataset,\n# batch_size=batch_size,\n# epochs=epochs,\n# steps_per_epoch=steps_per_epoch,\n# verbose=1,\n# callbacks=[] # \u4e0d\u4f7f\u7528\u56de\u8c03\u51fd\u6570\n# )\n#\n# # \u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\u6a21\u578b\n# score = model.evaluate(X_test_mapped, y_test, verbose=0)\n# print('\u6d4b\u8bd5\u635f\u5931:', score[0])\n# print('\u6d4b\u8bd5\u51c6\u786e\u7387:', score[1])\n#\n# mean_accuracy, fluctuation = calculate_accuracy_fluctuation(history)\n# print('\u9a8c\u8bc1\u96c6\u5e73\u5747\u51c6\u786e\u7387:', mean_accuracy)\n# print('\u51c6\u786e\u7387\u6ce2\u52a8:', fluctuation)\n#\n# plot_history(history)\n#\n# # \u5173\u95ed\u8f93\u51fa\u6587\u4ef6\u5e76\u6062\u590d\u6807\u51c6\u8f93\u51fa\n# sys.stdout = original_stdout\n# output_file.close()\n#\n# print(\"\u811a\u672c\u6267\u884c\u5b8c\u6bd5\u3002\u8bf7\u68c0\u67e5 'output_modified.log' \u4ee5\u83b7\u53d6\u6253\u5370\u8f93\u51fa\uff0c\u5e76\u5728\u5f53\u524d\u76ee\u5f55\u4e0b\u67e5\u770b\u4fdd\u5b58\u7684\u56fe\u8868\u3002\")\n\nimport sys\nimport os\nimport pennylane as qml\nfrom pennylane import numpy as np\nimport tensorflow as tf\nfrom sklearn.datasets import load_iris\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.preprocessing import OneHotEncoder, StandardScaler, MinMaxScaler\n\nimport matplotlib\nmatplotlib.use('Agg') # \u4f7f\u7528\u975e\u4ea4\u4e92\u5f0f\u540e\u7aef\uff0c\u4ee5\u4fbf\u5728\u6ca1\u6709\u663e\u793a\u73af\u5883\u7684\u60c5\u51b5\u4e0b\u8fd0\u884c\nimport matplotlib.pyplot as plt\n\nfrom tensorflow.keras.layers import Dense\nfrom importance_sampling.training import ConstantTimeImportanceTraining\nfrom importance_sampling.samplers import ConstantTimeSampler\nfrom importance_sampling.datasets import InMemoryDataset\nfrom importance_sampling.reweighting import BiasedReweightingPolicy\nfrom tensorflow.keras.callbacks import EarlyStopping\n\n# \u91cd\u5b9a\u5411\u6807\u51c6\u8f93\u51fa\u4ee5\u6355\u83b7\u6253\u5370\u8f93\u51fa\nclass Tee(object):\n def __init__(self, *files):\n self.files = files\n def write(self, obj):\n for f in self.files:\n f.write(obj)\n f.flush()\n def flush(self):\n for f in self.files:\n f.flush()\n\noutput_file = open('output_modified.log', 'w', encoding='utf-8')\noriginal_stdout = sys.stdout\nsys.stdout = Tee(sys.stdout, output_file)\n\ndef save_plot(fig, filename):\n fig.savefig(filename)\n plt.close(fig)\n\ndef plot_history(history):\n # \u7ed8\u5236\u8bad\u7ec3\u51c6\u786e\u7387\n fig, ax = plt.subplots(figsize=(10, 5))\n if 'accuracy' in history.history:\n ax.plot(history.history['accuracy'], label='Train Accuracy')\n elif 'acc' in history.history:\n ax.plot(history.history['acc'], label='Train Accuracy')\n else:\n print(\"No accuracy metric found in history.\")\n ax.set_title('Model Accuracy')\n ax.set_ylabel('Accuracy')\n ax.set_xlabel('Epoch')\n ax.legend(loc='upper left')\n save_plot(fig, 'accuracy_plot.png')\n\n # \u7ed8\u5236\u8bad\u7ec3\u635f\u5931\n fig, ax = plt.subplots(figsize=(10, 5))\n if 'loss' in history.history:\n ax.plot(history.history['loss'], label='Train Loss')\n else:\n print(\"No loss metric found in history.\")\n ax.set_title('Model Loss')\n ax.set_ylabel('Loss')\n ax.set_xlabel('Epoch')\n ax.legend(loc='upper right')\n save_plot(fig, 'loss_plot.png')\n\ndef calculate_accuracy_fluctuation(history):\n if 'accuracy' in history.history:\n accuracies = history.history['accuracy']\n elif 'acc' in history.history:\n accuracies = history.history['acc']\n else:\n print(\"No accuracy metric found in history.\")\n return None, None\n mean_accuracy = np.mean(accuracies)\n fluctuation = np.sum(np.abs(accuracies - mean_accuracy))\n return mean_accuracy, fluctuation\n\n# \u5b9a\u4e49\u5177\u67094\u4e2a\u91cf\u5b50\u6bd4\u7279\u7684\u91cf\u5b50\u8bbe\u5907\ndev = qml.device('default.qubit', wires=4)\n\n@qml.qnode(dev, interface='tf', diff_method='backprop')\ndef quantum_neural_network(inputs, weights):\n num_qubits = 4\n\n # \u5c06\u8f93\u5165\u6570\u636e\u7f16\u7801\u5230\u91cf\u5b50\u6001\u4e2d\n for i in range(num_qubits):\n qml.RY(inputs[i], wires=i)\n\n # \u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u7684\u53c2\u6570\u5316\u7ebf\u8def\n n_layers = weights.shape[0]\n for layer in range(n_layers):\n # \u5bf9\u6bcf\u4e2a\u91cf\u5b50\u6bd4\u7279\u5e94\u7528\u65cb\u8f6c\u95e8\n for i in range(num_qubits):\n params = weights[layer, i]\n qml.Rot(params[0], params[1], params[2], wires=i)\n # \u5e94\u7528\u7ea0\u7f20\u95e8\uff0c\u5982 CNOT \u95e8\n for i in range(num_qubits - 1):\n qml.CNOT(wires=[i, i + 1])\n\n # \u5728\u6bcf\u4e2a\u91cf\u5b50\u6bd4\u7279\u4e0a\u6d4b\u91cf PauliZ \u671f\u671b\u503c\n return [qml.expval(qml.PauliZ(i)) for i in range(num_qubits)]\n\n# \u5b9a\u4e49\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u5c42\nclass QuantumNeuralNetworkLayer(tf.keras.layers.Layer):\n def __init__(self, n_qubits, n_layers):\n super(QuantumNeuralNetworkLayer, self).__init__()\n self.n_qubits = n_qubits\n self.n_layers = n_layers\n\n # Initialize trainable weights with dtype=tf.float64\n init = tf.random_uniform_initializer(minval=0, maxval=2 * np.pi)\n self.q_weights = self.add_weight(\n shape=(n_layers, n_qubits, 3),\n initializer=init,\n trainable=True,\n name='q_weights',\n dtype=tf.float64\n )\n\n def call(self, inputs):\n def circuit(inputs):\n inputs = tf.cast(inputs, dtype=tf.float64)\n result = quantum_neural_network(inputs, self.q_weights)\n return result\n\n outputs = tf.map_fn(circuit, inputs, dtype=tf.float64)\n return outputs\n\n# \u5b9a\u4e49\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\nclass QuantumNeuralNetwork(tf.keras.Model):\n def __init__(self, n_qubits, n_layers, num_classes):\n super(QuantumNeuralNetwork, self).__init__()\n self.n_qubits = n_qubits\n self.n_layers = n_layers\n self.num_classes = num_classes\n\n self.quantum_layer = QuantumNeuralNetworkLayer(n_qubits, n_layers)\n self.output_layer = Dense(num_classes, activation='softmax', dtype=tf.float64)\n\n def call(self, inputs):\n q_output = self.quantum_layer(inputs)\n return self.output_layer(q_output)\n\ndef is_quantum_model(model):\n for layer in model.layers:\n if isinstance(layer, QuantumNeuralNetworkLayer):\n return True\n elif hasattr(layer, 'layers'):\n if is_quantum_model(layer):\n return True\n return False\n\nif __name__ == \"__main__\":\n # \u52a0\u8f7d\u6570\u636e\n iris = load_iris()\n X = iris['data'].astype(np.float64) # \u786e\u4fdd\u6570\u636e\u7c7b\u578b\u4e3a float64\n y = iris['target'].reshape(-1, 1)\n\n # \u5bf9\u6807\u7b7e\u8fdb\u884c One-Hot \u7f16\u7801\n encoder = OneHotEncoder(sparse_output=False)\n y_encoded = encoder.fit_transform(y).astype(np.float64)\n num_classes = y_encoded.shape[1]\n\n # \u76f4\u63a5\u62c6\u5206\u4e3a\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\n x_train, x_test, y_train, y_test = train_test_split(\n X, y_encoded, test_size=0.2, random_state=116, stratify=y_encoded\n )\n\n # \u6570\u636e\u6807\u51c6\u5316\u548c\u5f52\u4e00\u5316\n scaler = StandardScaler()\n X_train_standardized = scaler.fit_transform(x_train).astype(np.float64)\n X_test_standardized = scaler.transform(x_test).astype(np.float64)\n\n min_max_scaler = MinMaxScaler(feature_range=(0, np.pi))\n X_train_mapped = min_max_scaler.fit_transform(X_train_standardized).astype(np.float64)\n X_test_mapped = min_max_scaler.transform(X_test_standardized).astype(np.float64)\n\n # \u8d85\u53c2\u6570\n learning_rate = 0.02 # \u5b66\u4e60\u7387\n epochs = 100 # \u8bad\u7ec3\u8f6e\u6570\n batch_size = 16 # \u6279\u91cf\u5927\u5c0f\n n_qubits = 4 # 4 \u4e2a\u91cf\u5b50\u6bd4\u7279\n n_layers = 3 # \u5c42\u6570\n\n model = QuantumNeuralNetwork(\n n_qubits=n_qubits,\n n_layers=n_layers,\n num_classes=num_classes\n )\n\n if is_quantum_model(model):\n print(\"\u8be5\u6a21\u578b\u5305\u542b\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\uff08QNN\uff09\u5c42\u3002\")\n else:\n print(\"\u8be5\u6a21\u578b\u4e0d\u5305\u542b\u91cf\u5b50\u795e\u7ecf\u7f51\u7edc\uff08QNN\uff09\u5c42\u3002\")\n\n # \u7ed8\u5236\u91cf\u5b50\u7535\u8def\u56fe\n sample_inputs = np.ones(4, dtype=np.float64) * (np.pi / 2)\n sample_weights = np.zeros((n_layers, n_qubits, 3), dtype=np.float64)\n\n print(\"\\n\u91cf\u5b50\u7535\u8def\u56fe\uff1a\")\n print(qml.draw(quantum_neural_network)(sample_inputs, sample_weights))\n\n steps_per_epoch = len(X_train_mapped) // batch_size\n total_steps = epochs * steps_per_epoch\n\n # \u521b\u5efa\u6570\u636e\u96c6\uff08\u65e0\u9a8c\u8bc1\u96c6\uff09\n dataset = InMemoryDataset(\n X_train_mapped,\n y_train,\n None, # \u65e0\u9a8c\u8bc1\u96c6\u6570\u636e\n None,\n categorical=False\n )\n\n # \u521d\u59cb\u5316\u91cd\u52a0\u6743\u7b56\u7565\n reweighting_policy = BiasedReweightingPolicy(k=1.0)\n\n # \u521d\u59cb\u5316\u91cd\u8981\u6027\u91c7\u6837\u8bad\u7ec3\n wrapped_model = ConstantTimeImportanceTraining(model, score=\"loss\")\n\n # \u4fee\u6539\u91c7\u6837\u5668\u4ee5\u4f7f\u7528\u52a8\u6001\u7684 alpha \u548c beta\n def custom_sampler(dataset, batch_size, steps_per_epoch, epochs):\n return ConstantTimeSampler(\n dataset,\n reweighting_policy,\n model=wrapped_model.model,\n B=64,\n alpha=0.6,\n beta_start=0.4,\n beta_frames=total_steps\n )\n\n # \u5c06\u81ea\u5b9a\u4e49\u91c7\u6837\u5668\u8d4b\u7ed9 wrapped_model\n wrapped_model.sampler = custom_sampler\n\n # \u4f7f\u7528 per-sample \u7684\u635f\u5931\u51fd\u6570\n loss_fn = tf.keras.losses.CategoricalCrossentropy(from_logits=False, reduction=tf.keras.losses.Reduction.NONE)\n\n # \u7f16\u8bd1\u6a21\u578b\n model.compile(\n optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),\n loss=loss_fn,\n metrics=['accuracy'],\n run_eagerly=True # \u5e2e\u52a9\u8c03\u8bd5\n )\n\n # \u521b\u5efa\u4fdd\u5b58\u8bad\u7ec3\u6307\u6807\u7684\u56de\u8c03\u7c7b\n class MetricsHistory(tf.keras.callbacks.Callback):\n def on_train_begin(self, logs=None):\n self.history = {'loss': [], 'accuracy': []}\n\n def on_epoch_end(self, epoch, logs=None):\n self.history['loss'].append(logs.get('loss'))\n self.history['accuracy'].append(logs.get('accuracy'))\n\n # \u5b9e\u4f8b\u5316\u56de\u8c03\u51fd\u6570\n metrics_history = MetricsHistory()\n\n # \u4f7f\u7528 fit_dataset \u8bad\u7ec3\u6a21\u578b\n history = wrapped_model.fit_dataset(\n dataset=dataset,\n batch_size=batch_size,\n epochs=epochs,\n steps_per_epoch=steps_per_epoch,\n verbose=1,\n callbacks=[metrics_history] # \u4f7f\u7528\u81ea\u5b9a\u4e49\u56de\u8c03\u51fd\u6570\n )\n\n # \u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\u6a21\u578b\n score = model.evaluate(X_test_mapped, y_test, verbose=0)\n print('\u6d4b\u8bd5\u635f\u5931:', score[0])\n print('\u6d4b\u8bd5\u51c6\u786e\u7387:', score[1])\n\n # \u4f7f\u7528\u81ea\u5b9a\u4e49\u56de\u8c03\u51fd\u6570\u4e2d\u7684\u5386\u53f2\u8bb0\u5f55\u7ed8\u5236\u56fe\u8868\n plot_history(metrics_history)\n\n mean_accuracy, fluctuation = calculate_accuracy_fluctuation(metrics_history)\n if mean_accuracy is not None and fluctuation is not None:\n print('\u8bad\u7ec3\u96c6\u5e73\u5747\u51c6\u786e\u7387:', mean_accuracy)\n print('\u51c6\u786e\u7387\u6ce2\u52a8:', fluctuation)\n\n # \u5173\u95ed\u8f93\u51fa\u6587\u4ef6\u5e76\u6062\u590d\u6807\u51c6\u8f93\u51fa\n sys.stdout = original_stdout\n output_file.close()\n\n print(\"\u811a\u672c\u6267\u884c\u5b8c\u6bd5\u3002\u8bf7\u68c0\u67e5 'output_modified.log' \u4ee5\u83b7\u53d6\u6253\u5370\u8f93\u51fa\uff0c\u5e76\u5728\u5f53\u524d\u76ee\u5f55\u4e0b\u67e5\u770b\u4fdd\u5b58\u7684\u56fe\u8868\u3002\")\n\u91cd\u65b0\u5212\u5206\u4e0a\u8ff0\u4ee3\u7801\u7684\u6570\u636e\u96c6\uff0c\u5728\u8bad\u7ec3\u96c6\u4e2d\u5212\u5206\u4e00\u90e8\u5206\u4f5c\u4e3a\u9a8c\u8bc1\u96c6\uff0c\u987a\u4fbf\u68c0\u67e5\u4e00\u4e0b\u4e0a\u8ff0\u5185\u5bb9\u662f\u5426\u51fa\u73b0\u4e86\u6570\u636e\u6cc4\u9732"},{"assistant_initiates":false,"asst_text":"\u041a\u043e\u043d\u0435\u0447\u043d\u043e, \u044f \u043f\u0440\u043e\u0432\u0435\u0440\u044e \u043f\u0440\u0435\u0434\u043e\u0441\u0442\u0430\u0432\u043b\u0435\u043d\u043d\u044b\u0439 \u0432\u0430\u043c\u0438 \u043a\u043e\u0434 \u043d\u0430 \u043d\u0430\u043b\u0438\u0447\u0438\u0435 \u043e\u0448\u0438\u0431\u043e\u043a \u0438 \u043e\u0431\u044a\u0435\u0434\u0438\u043d\u044e \u0432\u0441\u0435 JavaScript \u0444\u0430\u0439\u043b\u044b \u0432 \u043e\u0434\u0438\u043d. \u041d\u0438\u0436\u0435 \u043f\u0440\u0438\u0432\u0435\u0434\u0435\u043d\u044b \u0448\u0430\u0433\u0438:\n\n1. **\u041f\u0440\u043e\u0432\u0435\u0440\u043a\u0430 \u043d\u0430 \u043e\u0448\u0438\u0431\u043a\u0438**\n2. **\u041e\u0431\u044a\u0435\u0434\u0438\u043d\u0435\u043d\u0438\u0435 \u0432\u0441\u0435\u0445 JS \u0444\u0430\u0439\u043b\u043e\u0432 \u0432 \u043e\u0434\u0438\u043d**\n\n## 1. \u041f\u0440\u043e\u0432\u0435\u0440\u043a\u0430 \u043d\u0430 \u043e\u0448\u0438\u0431\u043a\u0438\n\n### `package.json`\n\n\u0423\u0431\u0435\u0434\u0438\u0442\u0435\u0441\u044c, \u0447\u0442\u043e \u0432 `package.json` \u043f\u0440\u0438\u0441\u0443\u0442\u0441\u0442\u0432\u0443\u044e\u0442 \u0432\u0441\u0435 \u043d\u0435\u043e\u0431\u0445\u043e\u0434\u0438\u043c\u044b\u0435 \u0437\u0430\u0432\u0438\u0441\u0438\u043c\u043e\u0441\u0442\u0438:\n\n```json\n{\n \"name\": \"my-chat-app\",\n \"version\": \"0.0.1\",\n \"private\": true,\n \"scripts\": {\n \"android\": \"react-native run-android\",\n \"ios\": \"react-native run-ios\",\n \"start\": \"react-native start\",\n \"test\": \"jest\",\n \"lint\": \"eslint .\"\n },\n \"dependencies\": {\n \"@gradio/client\": \"^1.0.0\", // \u041f\u0440\u043e\u0432\u0435\u0440\u044c\u0442\u0435 \u0430\u043a\u0442\u0443\u0430\u043b\u044c\u043d\u0443\u044e \u0432\u0435\u0440\u0441\u0438\u044e\n \"react\": \"18.2.0\",\n \"react-native\": \"0.71.8\",\n \"react-native-gesture-handler\": \"^2.9.0\",\n \"react-native-reanimated\": \"^3.1.0\"\n },\n \"devDependencies\": {\n \"@babel/core\": \"^7.22.1\",\n \"@babel/runtime\": \"^7.22.0\",\n \"@react-native/eslint-config\": \"^0.71.0\",\n \"babel-jest\": \"^29.6.1\",\n \"eslint\": \"^8.43.0\",\n \"jest\": \"^29.6.1\",\n \"metro-react-native-babel-preset\": \"^0.76.12\",\n \"react-test-renderer\": \"18.2.0\"\n },\n \"jest\": {\n \"preset\": \"react-native\"\n }\n}\n```\n\n### `babel.config.js`\n\n\u0423\u0431\u0435\u0434\u0438\u0442\u0435\u0441\u044c, \u0447\u0442\u043e \u043a\u043e\u043d\u0444\u0438\u0433\u0443\u0440\u0430\u0446\u0438\u044f Babel \u043d\u0430\u0441\u0442\u0440\u043e\u0435\u043d\u0430 \u043f\u0440\u0430\u0432\u0438\u043b\u044c\u043d\u043e, \u043e\u0441\u043e\u0431\u0435\u043d\u043d\u043e \u0434\u043b\u044f `react-native-reanimated`:\n\n```javascript\nmodule.exports = {\n presets: ['module:metro-react-native-babel-preset'],\n plugins: [\n 'react-native-reanimated/plugin', // \u0414\u043e\u0431\u0430\u0432\u044c\u0442\u0435 \u044d\u0442\u043e\u0442 \u043f\u043b\u0430\u0433\u0438\u043d \u0434\u043b\u044f react-native-reanimated\n ],\n};\n```\n\n### \u041a\u043e\u0434\u044b \u043a\u043e\u043c\u043f\u043e\u043d\u0435\u043d\u0442\u043e\u0432 \u0438 \u0441\u0435\u0440\u0432\u0438\u0441\u043e\u0432\n\n\u041f\u0440\u043e\u0432\u0435\u0440\u0438\u043c \u0432\u0430\u0448\u0438 \u043a\u043e\u043c\u043f\u043e\u043d\u0435\u043d\u0442\u044b \u0438 \u0441\u0435\u0440\u0432\u0438\u0441\u044b \u043d\u0430 \u043d\u0430\u043b\u0438\u0447\u0438\u0435 \u043e\u0448\u0438\u0431\u043e\u043a.\n\n#### `src/services/api.js`\n\n```javascript\n// src/services/api.js\n\nimport { Client } from \"@gradio/client\";\n\nlet client = null;\n\nexport const connectClient = async () => {\n if (!client) {\n client = await Client.connect(\"yuntian-deng/o1mini\");\n }\n return client;\n};\n\nexport const enableInputs = async () => {\n const client = await connectClient();\n const result = await client.predict(\"/enable_inputs\", {});\n return result.data;\n};\n\nexport const resetTextbox = async () => {\n const client = await connectClient();\n const result = await client.predict(\"/reset_textbox\", {});\n return result.data;\n};\n\nexport const predict = async (inputText, top_p = 1, temperature = 1, chat_counter = 0, chatbot = []) => {\n const client = await connectClient();\n const result = await client.predict(\"/predict\", {\n inputs: inputText,\n top_p,\n temperature,\n chat_counter,\n chatbot,\n });\n return result.data;\n};\n\n// \u0415\u0441\u043b\u0438 \u043d\u0435\u043e\u0431\u0445\u043e\u0434\u0438\u043c\u044b \u0434\u043e\u043f\u043e\u043b\u043d\u0438\u0442\u0435\u043b\u044c\u043d\u044b\u0435 \u0444\u0443\u043d\u043a\u0446\u0438\u0438 \u0434\u043b\u044f /reset_textbox_1 \u0438 /predict_1, \u0434\u043e\u0431\u0430\u0432\u044c\u0442\u0435 \u0438\u0445 \u0430\u043d\u0430\u043b\u043e\u0433\u0438\u0447\u043d\u043e\n```\n\n**\u041e\u0448\u0438\u0431\u043a\u0438/\u041f\u0440\u0438\u043c\u0435\u0447\u0430\u043d\u0438\u044f:**\n- \u0423\u0431\u0435\u0434\u0438\u0442\u0435\u0441\u044c, \u0447\u0442\u043e \u043f\u0430\u043a\u0435\u0442 `@gradio/client` \u043f\u0440\u0430\u0432\u0438\u043b\u044c\u043d\u043e \u0443\u0441\u0442\u0430\u043d\u043e\u0432\u043b\u0435\u043d \u0438 \u0441\u043e\u0432\u043c\u0435\u0441\u0442\u0438\u043c \u0441 \u0432\u0430\u0448\u0435\u0439 \u0432\u0435\u0440\u0441\u0438\u0435\u0439 React Native.\n- \u041f\u0440\u043e\u0432\u0435\u0440\u044c\u0442\u0435 \u043f\u0440\u0430\u0432\u0438\u043b\u044c\u043d\u043e\u0441\u0442\u044c \u043c\u0430\u0440\u0448\u0440\u0443\u0442\u043e\u0432 API (\u043d\u0430\u043f\u0440\u0438\u043c\u0435\u0440, `/enable_inputs`, `/predict`).\n\n#### `src/components/ChatMessage.js`\n\n```javascript\n// src/components/ChatMessage.js\n\nimport React from 'react';\nimport { View, Text, StyleSheet } from 'react-native';\n\nconst ChatMessage = ({ message, isUser }) => {\n return (\n \n {message}\n \n );\n};\n\nconst styles = StyleSheet.create({\n messageContainer: {\n padding: 10,\n marginVertical: 5,\n borderRadius: 10,\n maxWidth: '80%',\n },\n userMessage: {\n backgroundColor: '#DCF8C6',\n alignSelf: 'flex-end',\n },\n botMessage: {\n backgroundColor: '#ECECEC',\n alignSelf: 'flex-start',\n },\n messageText: {\n fontSize: 16,\n },\n});\n\nexport default ChatMessage;\n```\n\n**\u041e\u0448\u0438\u0431\u043a\u0438/\u041f\u0440\u0438\u043c\u0435\u0447\u0430\u043d\u0438\u044f:**\n- \u041a\u043e\u043c\u043f\u043e\u043d\u0435\u043d\u0442 \u0432\u044b\u0433\u043b\u044f\u0434\u0438\u0442 \u043a\u043e\u0440\u0440\u0435\u043a\u0442\u043d\u043e.\n\n#### `src/components/ChatInput.js`\n\n```javascript\n// src/components/ChatInput.js\n\nimport React, { useState } from 'react';\nimport { View, TextInput, Button, StyleSheet } from 'react-native';\n\nconst ChatInput = ({ onSend }) => {\n const [text, setText] = useState('');\n\n const handleSend = () => {\n if (text.trim()) {\n onSend(text.trim());\n setText('');\n }\n };\n\n return (\n \n \n