examples/tutorial/VAELM.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "visible-blade",
   "metadata": {},
   "source": [
    "# VAELM\n",
    "\n",
    "可用的中文预训练参数：暂无"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "rotary-academy",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "v2.5.0\n"
     ]
    }
   ],
   "source": [
    "import uf\n",
    "\n",
    "print(uf.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "velvet-symbol",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "uf.VAELM(\n",
      "    vocab_file=\"../../ref/vocab.txt\",\n",
      "    max_seq_length=32,\n",
      "    init_checkpoint=None,\n",
      "    output_dir=None,\n",
      "    gpu_ids=None,\n",
      "    reduced_size=16,\n",
      "    topic_size=128,\n",
      "    hidden_size=64,\n",
      "    num_hidden_layers=1,\n",
      "    num_attention_heads=8,\n",
      "    do_lower_case=True,\n",
      "    truncate_method=\"LIFO\",\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "model = uf.VAELM(\n",
    "    \"../../ref/vocab.txt\",\n",
    "    max_seq_length=32,\n",
    "    reduced_size=16,            # VAE对于模型规格与数据体量的一致性，要求是极其严格的\n",
    "    topic_size=128,             # 这里由于训练数据较少，参数取值也设置较低\n",
    "    hidden_size=64,\n",
    "    num_hidden_layers=1,    \n",
    "    num_attention_heads=8,    \n",
    ")\n",
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "drawn-breathing",
   "metadata": {},
   "outputs": [],
   "source": [
    "X = [\n",
    "    \"天亮以前说再见，笑着泪流满面，去迎接应该你的，更好的明天\", \n",
    "    \"昙花若只一现，更要开的耀眼，别回头去拥有，属于你更好的世界\", \n",
    "    \"天亮以前说再见，让我留在今天，去保护我和你的，最好的夏天\", \n",
    "    \"情丝若水三千，只取一瓢眷恋，当你来过的纪念\",\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "vanilla-fault",
   "metadata": {},
   "source": [
    "# 训练"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "excellent-religion",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\tensorflow\\python\\util\\dispatch.py:1176: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\bert\\bert.py:272: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  query_layer = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\bert\\bert.py:282: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  key_layer = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\bert\\bert.py:292: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  value_layer = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\bert\\bert.py:416: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  attention_output = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\bert\\bert.py:428: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  intermediate_output = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\bert\\bert.py:438: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  layer_output = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\vae\\vae.py:103: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  transformer_output = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\vae\\vae.py:126: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  miu = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\vae\\vae.py:134: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  sigma = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\vae\\vae.py:157: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  decoder_input = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\vae\\vae.py:173: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  intermediate_output = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\vae\\vae.py:181: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  decoder_output = tf.layers.dense(\n",
      "c:\\Users\\Luv_d\\AppData\\Local\\Programs\\Python\\Python39\\lib\\site-packages\\uf-2.5.0-py3.9.egg\\uf\\apps\\vae\\vae.py:197: UserWarning: `tf.layers.dense` is deprecated and will be removed in a future version. Please use `tf.keras.layers.Dense` instead.\n",
      "  input_tensor = tf.layers.dense(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Build graph with 6,083,112 parameters (among which 2,027,704 are trainable)\n",
      "INFO:tensorflow:Running local_init_op\n",
      "INFO:tensorflow:Done running local_init_op\n",
      "INFO:tensorflow:Running training on 4 samples (step 0 -> 400)\n",
      "INFO:tensorflow:step 1, accuracy 0.0000, loss 9.949763, 1.24 steps/sec, 4.97 examples/sec\n",
      "INFO:tensorflow:step 4, accuracy 0.0000, loss 9.888888, 24.64 steps/sec, 98.56 examples/sec\n",
      "INFO:tensorflow:step 7, accuracy 0.0091, loss 9.826713, 21.20 steps/sec, 84.81 examples/sec\n",
      "INFO:tensorflow:step 9, accuracy 0.0091, loss 9.757114, 19.76 steps/sec, 79.04 examples/sec\n",
      "INFO:tensorflow:step 12, accuracy 0.0091, loss 9.665363, 23.60 steps/sec, 94.39 examples/sec\n",
      "INFO:tensorflow:step 14, accuracy 0.0091, loss 9.587683, 18.66 steps/sec, 74.65 examples/sec\n",
      "INFO:tensorflow:step 17, accuracy 0.0091, loss 9.502987, 24.03 steps/sec, 96.12 examples/sec\n",
      "INFO:tensorflow:step 20, accuracy 0.0364, loss 9.375195, 24.28 steps/sec, 97.13 examples/sec\n",
      "INFO:tensorflow:step 23, accuracy 0.0000, loss 9.257110, 24.18 steps/sec, 96.70 examples/sec\n",
      "INFO:tensorflow:step 26, accuracy 0.0182, loss 9.160007, 22.59 steps/sec, 90.34 examples/sec\n",
      "INFO:tensorflow:step 29, accuracy 0.0091, loss 9.043734, 23.15 steps/sec, 92.59 examples/sec\n",
      "INFO:tensorflow:step 32, accuracy 0.0000, loss 8.939093, 20.78 steps/sec, 83.12 examples/sec\n",
      "INFO:tensorflow:step 35, accuracy 0.0000, loss 8.832666, 21.69 steps/sec, 86.77 examples/sec\n",
      "INFO:tensorflow:step 38, accuracy 0.0000, loss 8.717691, 20.63 steps/sec, 82.51 examples/sec\n",
      "INFO:tensorflow:step 41, accuracy 0.0000, loss 8.599036, 23.01 steps/sec, 92.02 examples/sec\n",
      "INFO:tensorflow:step 44, accuracy 0.0000, loss 8.468253, 24.33 steps/sec, 97.31 examples/sec\n",
      "INFO:tensorflow:step 47, accuracy 0.0000, loss 8.341351, 24.53 steps/sec, 98.10 examples/sec\n",
      "INFO:tensorflow:step 50, accuracy 0.0000, loss 8.208193, 24.69 steps/sec, 98.75 examples/sec\n",
      "INFO:tensorflow:step 53, accuracy 0.0000, loss 8.066173, 25.58 steps/sec, 102.31 examples/sec\n",
      "INFO:tensorflow:step 56, accuracy 0.0000, loss 7.915408, 23.78 steps/sec, 95.14 examples/sec\n",
      "INFO:tensorflow:step 59, accuracy 0.0000, loss 7.769206, 24.05 steps/sec, 96.19 examples/sec\n",
      "INFO:tensorflow:step 62, accuracy 0.0000, loss 7.606905, 23.09 steps/sec, 92.36 examples/sec\n",
      "INFO:tensorflow:step 65, accuracy 0.0000, loss 7.446529, 21.73 steps/sec, 86.91 examples/sec\n",
      "INFO:tensorflow:step 68, accuracy 0.0000, loss 7.270027, 20.96 steps/sec, 83.83 examples/sec\n",
      "INFO:tensorflow:step 71, accuracy 0.0000, loss 7.102561, 20.85 steps/sec, 83.39 examples/sec\n",
      "INFO:tensorflow:step 74, accuracy 0.0000, loss 6.919677, 21.08 steps/sec, 84.33 examples/sec\n",
      "INFO:tensorflow:step 77, accuracy 0.0000, loss 6.738069, 23.23 steps/sec, 92.93 examples/sec\n",
      "INFO:tensorflow:step 80, accuracy 0.0000, loss 6.552164, 25.65 steps/sec, 102.60 examples/sec\n",
      "INFO:tensorflow:step 83, accuracy 0.0000, loss 6.355633, 25.70 steps/sec, 102.79 examples/sec\n",
      "INFO:tensorflow:step 86, accuracy 0.0000, loss 6.167645, 25.21 steps/sec, 100.84 examples/sec\n",
      "INFO:tensorflow:step 89, accuracy 0.0000, loss 5.967668, 24.59 steps/sec, 98.34 examples/sec\n",
      "INFO:tensorflow:step 92, accuracy 0.0000, loss 5.768925, 24.68 steps/sec, 98.74 examples/sec\n",
      "INFO:tensorflow:step 95, accuracy 0.0000, loss 5.570368, 24.65 steps/sec, 98.59 examples/sec\n",
      "INFO:tensorflow:step 98, accuracy 0.0000, loss 5.374008, 25.88 steps/sec, 103.53 examples/sec\n",
      "INFO:tensorflow:step 101, accuracy 0.0000, loss 5.181190, 25.78 steps/sec, 103.12 examples/sec\n",
      "INFO:tensorflow:step 104, accuracy 0.0000, loss 4.996796, 25.87 steps/sec, 103.50 examples/sec\n",
      "INFO:tensorflow:step 107, accuracy 0.0000, loss 4.823739, 24.76 steps/sec, 99.02 examples/sec\n",
      "INFO:tensorflow:step 110, accuracy 0.0000, loss 4.660326, 25.77 steps/sec, 103.07 examples/sec\n",
      "INFO:tensorflow:step 113, accuracy 0.0000, loss 4.508484, 25.24 steps/sec, 100.95 examples/sec\n",
      "INFO:tensorflow:step 116, accuracy 0.0000, loss 4.379753, 24.82 steps/sec, 99.28 examples/sec\n",
      "INFO:tensorflow:step 119, accuracy 0.0000, loss 4.254757, 25.00 steps/sec, 100.02 examples/sec\n",
      "INFO:tensorflow:step 122, accuracy 0.0000, loss 4.154023, 24.86 steps/sec, 99.46 examples/sec\n",
      "INFO:tensorflow:step 125, accuracy 0.0000, loss 4.059400, 23.79 steps/sec, 95.15 examples/sec\n",
      "INFO:tensorflow:step 128, accuracy 0.0000, loss 3.989591, 21.48 steps/sec, 85.93 examples/sec\n",
      "INFO:tensorflow:step 131, accuracy 0.0000, loss 3.930040, 23.55 steps/sec, 94.22 examples/sec\n",
      "INFO:tensorflow:step 134, accuracy 0.0000, loss 3.896353, 24.63 steps/sec, 98.50 examples/sec\n",
      "INFO:tensorflow:step 137, accuracy 0.0545, loss 3.834077, 22.82 steps/sec, 91.29 examples/sec\n",
      "INFO:tensorflow:step 140, accuracy 0.0364, loss 3.817227, 22.22 steps/sec, 88.88 examples/sec\n",
      "INFO:tensorflow:step 143, accuracy 0.0636, loss 3.772589, 25.55 steps/sec, 102.21 examples/sec\n",
      "INFO:tensorflow:step 146, accuracy 0.0909, loss 3.689115, 21.04 steps/sec, 84.17 examples/sec\n",
      "INFO:tensorflow:step 149, accuracy 0.0909, loss 3.651202, 24.08 steps/sec, 96.32 examples/sec\n",
      "INFO:tensorflow:step 152, accuracy 0.0909, loss 3.634161, 24.98 steps/sec, 99.93 examples/sec\n",
      "INFO:tensorflow:step 155, accuracy 0.1000, loss 3.505689, 21.98 steps/sec, 87.94 examples/sec\n",
      "INFO:tensorflow:step 158, accuracy 0.1000, loss 3.479036, 20.57 steps/sec, 82.29 examples/sec\n",
      "INFO:tensorflow:step 161, accuracy 0.1000, loss 3.365995, 20.42 steps/sec, 81.66 examples/sec\n",
      "INFO:tensorflow:step 164, accuracy 0.1000, loss 3.334129, 22.23 steps/sec, 88.94 examples/sec\n",
      "INFO:tensorflow:step 167, accuracy 0.1000, loss 3.110067, 21.65 steps/sec, 86.60 examples/sec\n",
      "INFO:tensorflow:step 170, accuracy 0.1364, loss 3.074591, 21.84 steps/sec, 87.36 examples/sec\n",
      "INFO:tensorflow:step 173, accuracy 0.1636, loss 2.899144, 22.15 steps/sec, 88.59 examples/sec\n",
      "INFO:tensorflow:step 176, accuracy 0.2091, loss 2.784977, 22.14 steps/sec, 88.55 examples/sec\n",
      "INFO:tensorflow:step 179, accuracy 0.1727, loss 2.764922, 22.47 steps/sec, 89.89 examples/sec\n",
      "INFO:tensorflow:step 182, accuracy 0.2182, loss 2.504732, 22.38 steps/sec, 89.53 examples/sec\n",
      "INFO:tensorflow:step 185, accuracy 0.2364, loss 2.309060, 21.88 steps/sec, 87.53 examples/sec\n",
      "INFO:tensorflow:step 188, accuracy 0.2636, loss 2.083245, 24.82 steps/sec, 99.29 examples/sec\n",
      "INFO:tensorflow:step 191, accuracy 0.2636, loss 2.062997, 25.69 steps/sec, 102.76 examples/sec\n",
      "INFO:tensorflow:step 194, accuracy 0.3909, loss 1.971056, 25.39 steps/sec, 101.54 examples/sec\n",
      "INFO:tensorflow:step 197, accuracy 0.3909, loss 1.754604, 26.37 steps/sec, 105.46 examples/sec\n",
      "INFO:tensorflow:step 200, accuracy 0.4091, loss 1.534916, 25.68 steps/sec, 102.73 examples/sec\n",
      "INFO:tensorflow:step 203, accuracy 0.4909, loss 1.290973, 20.89 steps/sec, 83.57 examples/sec\n",
      "INFO:tensorflow:step 206, accuracy 0.5182, loss 1.275999, 24.53 steps/sec, 98.12 examples/sec\n",
      "INFO:tensorflow:step 209, accuracy 0.4727, loss 1.273349, 25.66 steps/sec, 102.63 examples/sec\n",
      "INFO:tensorflow:step 212, accuracy 0.5545, loss 1.193465, 22.04 steps/sec, 88.15 examples/sec\n",
      "INFO:tensorflow:step 215, accuracy 0.5364, loss 1.220489, 24.84 steps/sec, 99.36 examples/sec\n",
      "INFO:tensorflow:step 218, accuracy 0.5727, loss 0.988680, 25.38 steps/sec, 101.53 examples/sec\n",
      "INFO:tensorflow:step 221, accuracy 0.6636, loss 0.868106, 23.53 steps/sec, 94.10 examples/sec\n",
      "INFO:tensorflow:step 224, accuracy 0.7273, loss 0.688851, 22.74 steps/sec, 90.97 examples/sec\n",
      "INFO:tensorflow:step 227, accuracy 0.6455, loss 0.738063, 25.54 steps/sec, 102.17 examples/sec\n",
      "INFO:tensorflow:step 230, accuracy 0.6909, loss 0.739638, 26.41 steps/sec, 105.64 examples/sec\n",
      "INFO:tensorflow:step 233, accuracy 0.6909, loss 0.691240, 25.99 steps/sec, 103.95 examples/sec\n",
      "INFO:tensorflow:step 236, accuracy 0.7000, loss 0.762913, 23.18 steps/sec, 92.72 examples/sec\n",
      "INFO:tensorflow:step 239, accuracy 0.7909, loss 0.537997, 24.86 steps/sec, 99.46 examples/sec\n",
      "INFO:tensorflow:step 242, accuracy 0.6818, loss 0.831308, 26.08 steps/sec, 104.32 examples/sec\n",
      "INFO:tensorflow:step 245, accuracy 0.7455, loss 0.651303, 26.34 steps/sec, 105.35 examples/sec\n",
      "INFO:tensorflow:step 248, accuracy 0.7364, loss 0.631732, 25.63 steps/sec, 102.52 examples/sec\n",
      "INFO:tensorflow:step 251, accuracy 0.7909, loss 0.561091, 26.30 steps/sec, 105.19 examples/sec\n",
      "INFO:tensorflow:step 254, accuracy 0.6909, loss 0.773264, 24.85 steps/sec, 99.41 examples/sec\n",
      "INFO:tensorflow:step 257, accuracy 0.7909, loss 0.426662, 25.34 steps/sec, 101.38 examples/sec\n",
      "INFO:tensorflow:step 260, accuracy 0.7636, loss 0.590151, 23.55 steps/sec, 94.21 examples/sec\n",
      "INFO:tensorflow:step 263, accuracy 0.8091, loss 0.439886, 25.43 steps/sec, 101.71 examples/sec\n",
      "INFO:tensorflow:step 266, accuracy 0.8000, loss 0.460066, 26.65 steps/sec, 106.61 examples/sec\n",
      "INFO:tensorflow:step 269, accuracy 0.7636, loss 0.536575, 26.20 steps/sec, 104.78 examples/sec\n",
      "INFO:tensorflow:step 272, accuracy 0.7727, loss 0.468719, 26.56 steps/sec, 106.24 examples/sec\n",
      "INFO:tensorflow:step 275, accuracy 0.8273, loss 0.359817, 26.11 steps/sec, 104.43 examples/sec\n",
      "INFO:tensorflow:step 278, accuracy 0.8000, loss 0.353747, 26.23 steps/sec, 104.91 examples/sec\n",
      "INFO:tensorflow:step 281, accuracy 0.8273, loss 0.384860, 26.09 steps/sec, 104.36 examples/sec\n",
      "INFO:tensorflow:step 284, accuracy 0.8091, loss 0.455236, 26.31 steps/sec, 105.22 examples/sec\n",
      "INFO:tensorflow:step 287, accuracy 0.8000, loss 0.444382, 25.06 steps/sec, 100.25 examples/sec\n",
      "INFO:tensorflow:step 290, accuracy 0.8182, loss 0.394423, 25.42 steps/sec, 101.67 examples/sec\n",
      "INFO:tensorflow:step 293, accuracy 0.8455, loss 0.251454, 25.87 steps/sec, 103.49 examples/sec\n",
      "INFO:tensorflow:step 296, accuracy 0.8364, loss 0.375705, 25.59 steps/sec, 102.37 examples/sec\n",
      "INFO:tensorflow:step 299, accuracy 0.7909, loss 0.446858, 26.57 steps/sec, 106.29 examples/sec\n",
      "INFO:tensorflow:step 302, accuracy 0.8273, loss 0.361514, 25.36 steps/sec, 101.45 examples/sec\n",
      "INFO:tensorflow:step 305, accuracy 0.8091, loss 0.408950, 25.98 steps/sec, 103.91 examples/sec\n",
      "INFO:tensorflow:step 308, accuracy 0.7909, loss 0.356028, 26.24 steps/sec, 104.96 examples/sec\n",
      "INFO:tensorflow:step 311, accuracy 0.9000, loss 0.264354, 26.05 steps/sec, 104.20 examples/sec\n",
      "INFO:tensorflow:step 314, accuracy 0.8727, loss 0.245719, 26.46 steps/sec, 105.86 examples/sec\n",
      "INFO:tensorflow:step 317, accuracy 0.8727, loss 0.226581, 25.90 steps/sec, 103.61 examples/sec\n",
      "INFO:tensorflow:step 320, accuracy 0.8727, loss 0.370148, 26.90 steps/sec, 107.60 examples/sec\n",
      "INFO:tensorflow:step 323, accuracy 0.7818, loss 0.352911, 25.89 steps/sec, 103.55 examples/sec\n",
      "INFO:tensorflow:step 326, accuracy 0.8273, loss 0.358762, 25.21 steps/sec, 100.83 examples/sec\n",
      "INFO:tensorflow:step 329, accuracy 0.8091, loss 0.348968, 25.46 steps/sec, 101.84 examples/sec\n",
      "INFO:tensorflow:step 332, accuracy 0.8000, loss 0.361838, 25.36 steps/sec, 101.42 examples/sec\n",
      "INFO:tensorflow:step 335, accuracy 0.8636, loss 0.294465, 25.86 steps/sec, 103.45 examples/sec\n",
      "INFO:tensorflow:step 338, accuracy 0.8455, loss 0.308141, 25.92 steps/sec, 103.68 examples/sec\n",
      "INFO:tensorflow:step 341, accuracy 0.8727, loss 0.259370, 26.71 steps/sec, 106.84 examples/sec\n",
      "INFO:tensorflow:step 344, accuracy 0.8727, loss 0.184372, 26.02 steps/sec, 104.10 examples/sec\n",
      "INFO:tensorflow:step 347, accuracy 0.8455, loss 0.388385, 26.62 steps/sec, 106.49 examples/sec\n",
      "INFO:tensorflow:step 350, accuracy 0.8818, loss 0.271496, 25.84 steps/sec, 103.36 examples/sec\n",
      "INFO:tensorflow:step 353, accuracy 0.8909, loss 0.206560, 25.91 steps/sec, 103.66 examples/sec\n",
      "INFO:tensorflow:step 356, accuracy 0.8455, loss 0.342267, 25.83 steps/sec, 103.33 examples/sec\n",
      "INFO:tensorflow:step 359, accuracy 0.8364, loss 0.284052, 24.95 steps/sec, 99.79 examples/sec\n",
      "INFO:tensorflow:step 362, accuracy 0.8182, loss 0.555393, 24.63 steps/sec, 98.50 examples/sec\n",
      "INFO:tensorflow:step 365, accuracy 0.9000, loss 0.244837, 25.06 steps/sec, 100.26 examples/sec\n",
      "INFO:tensorflow:step 368, accuracy 0.9091, loss 0.166558, 25.88 steps/sec, 103.52 examples/sec\n",
      "INFO:tensorflow:step 371, accuracy 0.8364, loss 0.334464, 26.07 steps/sec, 104.28 examples/sec\n",
      "INFO:tensorflow:step 374, accuracy 0.8909, loss 0.182515, 25.32 steps/sec, 101.29 examples/sec\n",
      "INFO:tensorflow:step 377, accuracy 0.9091, loss 0.229183, 25.17 steps/sec, 100.70 examples/sec\n",
      "INFO:tensorflow:step 380, accuracy 0.8818, loss 0.185641, 24.49 steps/sec, 97.94 examples/sec\n",
      "INFO:tensorflow:step 383, accuracy 0.8909, loss 0.251940, 26.15 steps/sec, 104.58 examples/sec\n",
      "INFO:tensorflow:step 386, accuracy 0.8455, loss 0.330600, 26.04 steps/sec, 104.16 examples/sec\n",
      "INFO:tensorflow:step 389, accuracy 0.8636, loss 0.282349, 25.64 steps/sec, 102.56 examples/sec\n",
      "INFO:tensorflow:step 392, accuracy 0.8818, loss 0.363174, 25.71 steps/sec, 102.84 examples/sec\n",
      "INFO:tensorflow:step 395, accuracy 0.9091, loss 0.194383, 24.18 steps/sec, 96.71 examples/sec\n",
      "INFO:tensorflow:step 398, accuracy 0.9273, loss 0.221734, 24.39 steps/sec, 97.56 examples/sec\n",
      "INFO:tensorflow:step 400, accuracy 0.9000, loss 0.187470, 25.25 steps/sec, 101.01 examples/sec\n"
     ]
    }
   ],
   "source": [
    "model.fit(X, total_steps=400, learning_rate=0.01, optimizer=\"lamb\")        # 模型较小，可以适当提高学习率"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fallen-hundred",
   "metadata": {},
   "source": [
    "# 推理"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "opposite-pantyhose",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Running inference on 4 samples\n",
      "INFO:tensorflow:process 100.0%, 25.24 examples/sec\n",
      "{'miu': array([[-0.4396022 ,  0.53117496,  0.17832753,  0.3916682 ,  0.5434318 ,\n",
      "        -0.349801  ,  0.4876728 ,  0.06779953,  0.16947743, -0.2487141 ,\n",
      "        -0.47011247, -0.35064802,  0.38101462, -0.39234826,  0.76904225,\n",
      "         0.7980169 , -0.31436703, -0.6142652 , -0.33549577, -0.33714166,\n",
      "         0.73015195,  0.47610235,  0.00210417, -0.09950063,  0.18604045,\n",
      "        -0.11020187, -0.48238188,  0.14054313, -0.51977366, -0.6530189 ,\n",
      "         0.29201305, -0.11135881, -0.6522274 , -0.01328439, -0.6988805 ,\n",
      "        -0.1957497 , -0.61858803,  0.4604739 , -0.16431828, -0.64519733,\n",
      "        -0.28615648, -0.10812853,  0.8175854 , -0.15867183,  0.23407696,\n",
      "         0.25935042,  0.48445275,  0.7359865 , -0.4633496 ,  0.03561433,\n",
      "        -0.00894808,  0.34221044,  0.46491835, -0.5450153 ,  0.3270287 ,\n",
      "        -0.4716175 , -0.03313906,  0.11883581, -0.28118053,  0.53394926,\n",
      "         0.47560224, -0.7123323 , -0.28942183,  0.01549871, -0.6245871 ,\n",
      "         0.11586095,  0.3069579 , -0.09454164,  0.24998963, -0.49753317,\n",
      "        -0.36490378, -0.0467333 ,  0.16291521,  0.03164184, -0.02466292,\n",
      "         0.77220887,  0.08216386, -0.39379105,  0.22965193, -0.6106302 ,\n",
      "         0.12905608,  0.03143375,  0.2750468 , -0.11824985,  0.11386785,\n",
      "        -0.21679497, -0.30680424,  0.15367134, -0.23380323, -0.645449  ,\n",
      "         0.54022264,  0.7515234 ,  0.17085446, -0.32974073, -0.5721658 ,\n",
      "         0.45940095, -0.2601403 ,  0.08977263,  0.01357785,  0.28882757,\n",
      "         0.02907572, -0.54030764,  0.23230802, -0.16057894,  0.8285682 ,\n",
      "        -0.21954456, -0.23607539, -0.3354532 ,  0.6094573 ,  0.01860696,\n",
      "        -0.3252273 , -0.06578828, -0.82433325,  0.16053909,  0.11647375,\n",
      "        -0.4680446 ,  0.41359082,  0.10610711, -0.05752654, -0.82179433,\n",
      "        -0.4556519 ,  0.15392034,  0.40726206,  0.22530027, -0.2509305 ,\n",
      "         0.05658961, -0.01185258, -0.3808937 ],\n",
      "       [-0.40424427, -0.6613178 , -0.55455583, -0.54453754,  0.15113   ,\n",
      "        -0.7264001 , -0.1964842 , -0.12757275,  0.3063119 , -0.09371693,\n",
      "        -0.78356874, -0.177179  , -0.61630243,  0.57635134,  0.02743418,\n",
      "        -0.6013617 ,  0.7516876 , -0.51784784, -0.62055796, -0.40939158,\n",
      "        -0.17829244,  0.13516863, -0.08669522,  0.4725463 ,  0.00599788,\n",
      "        -0.16556753, -0.4814041 , -0.09986823,  0.3095039 , -0.8197245 ,\n",
      "         0.35185435, -0.76147985,  0.22937186,  0.18708481,  0.29995713,\n",
      "         0.15827122,  0.02635231,  0.27391377,  0.5452139 ,  0.3958741 ,\n",
      "        -0.3562757 ,  0.0354724 ,  0.05675625, -0.4242439 , -0.02828944,\n",
      "        -0.56530756,  0.16176453, -0.31689498,  0.01594808,  0.47264296,\n",
      "         0.57770824,  0.14317366, -0.20788841,  0.12255213, -0.34458688,\n",
      "        -0.05454639,  0.61516744, -0.67166615,  0.7109099 ,  0.06450709,\n",
      "         0.13677163, -0.25293982, -0.4495866 ,  0.39292803, -0.08624236,\n",
      "        -0.42750838,  0.40305406,  0.5100008 , -0.5650745 , -0.39525053,\n",
      "        -0.29834765,  0.32661283,  0.38759992, -0.4559074 ,  0.75325495,\n",
      "         0.31332484, -0.07387909,  0.52348846,  0.60346425,  0.25267062,\n",
      "         0.59518296, -0.06666263,  0.14753018, -0.1554321 ,  0.39860487,\n",
      "        -0.5016212 , -0.20139454, -0.6015193 ,  0.8241274 , -0.4132846 ,\n",
      "         0.22998306,  0.4072479 , -0.7947555 , -0.2875811 , -0.49485567,\n",
      "        -0.07750013,  0.3428134 ,  0.50484043,  0.70465666, -0.5235935 ,\n",
      "         0.22626197,  0.16179188,  0.12681942, -0.2762881 ,  0.8757036 ,\n",
      "         0.6251476 , -0.34778562, -0.5337127 , -0.23631097, -0.07493189,\n",
      "         0.13019995,  0.37212723, -0.15985766,  0.35012403, -0.17318068,\n",
      "        -0.04319626,  0.6262446 , -0.7358355 ,  0.8688896 , -0.6850959 ,\n",
      "         0.15048127,  0.56519634,  0.3139976 , -0.7050873 , -0.54715115,\n",
      "         0.77197987, -0.4063307 , -0.56046206],\n",
      "       [-0.3642728 ,  0.46004498,  0.01059638,  0.32406574,  0.53835565,\n",
      "        -0.35064748,  0.45198888,  0.07451502,  0.01790816, -0.21773584,\n",
      "        -0.4787502 , -0.38566804,  0.17716734, -0.23125465,  0.6740272 ,\n",
      "         0.7275375 , -0.21396048, -0.6463626 , -0.48730168, -0.48037007,\n",
      "         0.6404515 ,  0.36057398, -0.01316697,  0.09231912,  0.29894668,\n",
      "        -0.0330594 , -0.6024469 ,  0.07630484, -0.38744673, -0.667308  ,\n",
      "         0.26080528, -0.1799443 , -0.50601375, -0.03865441, -0.58848375,\n",
      "        -0.178581  , -0.52887404,  0.36266178,  0.05854804, -0.51894945,\n",
      "        -0.35518825,  0.00790745,  0.72956234, -0.29058093,  0.0846502 ,\n",
      "         0.08107164,  0.5253183 ,  0.65459776, -0.3534064 ,  0.13854381,\n",
      "        -0.03686604,  0.39595297,  0.47555503, -0.35170725,  0.18984316,\n",
      "        -0.3834523 ,  0.10976946,  0.07115378, -0.12964062,  0.49872229,\n",
      "         0.5209197 , -0.5893627 , -0.38104895,  0.07993336, -0.52138287,\n",
      "        -0.10246953,  0.35107628, -0.09562219,  0.20706494, -0.41679892,\n",
      "        -0.45118335,  0.07654575,  0.05550493, -0.05588711,  0.16146721,\n",
      "         0.6882174 , -0.06010684, -0.30950406,  0.36571613, -0.510865  ,\n",
      "         0.03544237, -0.03008462,  0.36853167, -0.06452294,  0.15827432,\n",
      "        -0.23604253, -0.2281504 , -0.0185689 , -0.20900336, -0.57493144,\n",
      "         0.427282  ,  0.7825432 ,  0.07771046, -0.429275  , -0.64938486,\n",
      "         0.2931583 , -0.2670598 , -0.08133707,  0.05289136,  0.06665377,\n",
      "        -0.06810778, -0.51975286,  0.24316153, -0.04403505,  0.84778327,\n",
      "        -0.16769047, -0.15489088, -0.38689238,  0.51435137,  0.1327065 ,\n",
      "        -0.38551414, -0.11884423, -0.77674544,  0.1371838 ,  0.02346713,\n",
      "        -0.43203786,  0.49232924, -0.08234139, -0.02754244, -0.8178904 ,\n",
      "        -0.4164275 ,  0.32128194,  0.35656524,  0.21968696, -0.22027534,\n",
      "         0.101785  , -0.23617807, -0.3583702 ],\n",
      "       [ 0.51699   , -0.05625712, -0.46980837,  0.1681818 ,  0.2717811 ,\n",
      "        -0.24365702,  0.23892926, -0.4320021 , -0.7521961 ,  0.05039456,\n",
      "        -0.58603185, -0.19359787, -0.52544624,  0.3090927 , -0.5597222 ,\n",
      "        -0.10307791,  0.01066909, -0.4828461 , -0.5633853 , -0.7958153 ,\n",
      "        -0.50360787, -0.25945395, -0.36204723,  0.7365016 ,  0.74419653,\n",
      "         0.33362523, -0.69249564, -0.47149196,  0.53172225, -0.5156852 ,\n",
      "        -0.26261383,  0.04042988,  0.71059215, -0.18021889,  0.5622784 ,\n",
      "         0.08040801,  0.41929275, -0.48548552,  0.5546837 ,  0.484002  ,\n",
      "        -0.10581797,  0.19771634, -0.6374899 , -0.8068022 , -0.6890977 ,\n",
      "        -0.63238823,  0.54130673, -0.17360899,  0.43616864,  0.4141887 ,\n",
      "        -0.5260514 ,  0.53272367,  0.4639936 ,  0.77233076, -0.50562793,\n",
      "         0.14527456,  0.52342486,  0.3969197 ,  0.3626948 , -0.04184609,\n",
      "         0.5647089 ,  0.72535104, -0.5060267 ,  0.13688307,  0.4256356 ,\n",
      "        -0.64741695, -0.05520255, -0.33446285,  0.0542817 ,  0.484301  ,\n",
      "        -0.62048656,  0.3161051 , -0.7106597 , -0.40326267,  0.6463302 ,\n",
      "        -0.29725516, -0.50547385,  0.25041732,  0.5518853 ,  0.69279885,\n",
      "        -0.6185073 , -0.47554046,  0.6717064 ,  0.25623134,  0.2763538 ,\n",
      "        -0.08276364, -0.32833782, -0.6685877 , -0.52612495,  0.01979494,\n",
      "        -0.44114685,  0.57690847, -0.19277796, -0.6443098 , -0.65701175,\n",
      "        -0.7466271 , -0.0226044 , -0.73293847, -0.29825896, -0.4849551 ,\n",
      "        -0.4149438 , -0.3612215 ,  0.3061612 ,  0.26462027,  0.6999105 ,\n",
      "         0.03967398,  0.5756744 , -0.31545925,  0.03272838,  0.51458156,\n",
      "        -0.6775815 , -0.19910872,  0.298664  , -0.02839263, -0.24655114,\n",
      "        -0.43876162,  0.33000016, -0.5196884 , -0.20472391, -0.03096777,\n",
      "        -0.22015893,  0.5531062 , -0.4260221 ,  0.46003515,  0.23697607,\n",
      "        -0.20001571, -0.780906  ,  0.08472728]], dtype=float32), 'sigma': array([[-0.3831867 , -0.52395594, -0.39490977, -0.33780122, -0.41583732,\n",
      "        -0.47149825, -0.40949988, -0.34398192, -0.32951543, -0.31029698,\n",
      "        -0.54492164, -0.41599876, -0.4324875 , -0.3431378 , -0.45812592,\n",
      "        -0.25838342, -0.45791572, -0.42541614, -0.40994573, -0.48496142,\n",
      "        -0.37105954, -0.2794265 , -0.37017506, -0.55627453, -0.35677972,\n",
      "        -0.37343624, -0.383798  , -0.38137573, -0.40318215, -0.5163089 ,\n",
      "        -0.32750708, -0.42829078, -0.5289174 , -0.33509782, -0.49119   ,\n",
      "        -0.38223416, -0.41587228, -0.46118608, -0.52699506, -0.4486439 ,\n",
      "        -0.35629532, -0.33426616, -0.68044096, -0.5730364 , -0.2363012 ,\n",
      "        -0.63126826, -0.45245185, -0.4358171 , -0.44235292, -0.43667004,\n",
      "        -0.4293162 , -0.35206518, -0.3358361 , -0.5113451 , -0.2972256 ,\n",
      "        -0.36040854, -0.52424973, -0.49219388, -0.43922335, -0.32743767,\n",
      "        -0.40915126, -0.4421557 , -0.3465399 , -0.24146679, -0.3643712 ,\n",
      "        -0.4596647 , -0.3304459 , -0.4692658 , -0.29292592, -0.3044884 ,\n",
      "        -0.37971488, -0.38895005, -0.3393326 , -0.47503147, -0.3558076 ,\n",
      "        -0.3351344 , -0.2907724 , -0.31920582, -0.42375913, -0.5482773 ,\n",
      "        -0.3652358 , -0.24286714, -0.44433093, -0.39158028, -0.36590925,\n",
      "        -0.29350862, -0.34788963, -0.38077998, -0.47669187, -0.3440472 ,\n",
      "        -0.36944002, -0.52661586, -0.44760892, -0.4727469 , -0.54288644,\n",
      "        -0.42217866, -0.2958092 , -0.41930237, -0.507913  , -0.39832836,\n",
      "        -0.33301958, -0.48628405, -0.31339568, -0.3857982 , -0.40353784,\n",
      "        -0.3976823 , -0.4355631 , -0.25579527, -0.47944912, -0.33822495,\n",
      "        -0.42450094, -0.29456958, -0.4118774 , -0.43760365, -0.29186782,\n",
      "        -0.19793081, -0.43110508, -0.4421649 , -0.45649183, -0.30406314,\n",
      "        -0.32333088, -0.40880707, -0.38601798, -0.3375841 , -0.35035115,\n",
      "        -0.47714743, -0.45460117, -0.30579388],\n",
      "       [-0.45851618, -0.8126475 , -0.5778383 , -0.4340984 , -0.38926283,\n",
      "        -0.6052165 , -0.504553  , -0.48874742, -0.44569477, -0.3518301 ,\n",
      "        -0.58542645, -0.42095643, -0.6020549 , -0.62612903, -0.54076815,\n",
      "        -0.6840471 , -0.41007262, -0.6015291 , -0.47547355, -0.34628716,\n",
      "        -0.5558731 , -0.38952506, -0.45283544, -0.53094584, -0.41673958,\n",
      "        -0.5672391 , -0.5418804 , -0.3373803 , -0.6062407 , -0.5406409 ,\n",
      "        -0.3945115 , -0.46906376, -0.6198423 , -0.40681604, -0.5748803 ,\n",
      "        -0.5177635 , -0.5943926 , -0.611834  , -0.6578959 , -0.73868144,\n",
      "        -0.3899627 , -0.3567461 , -0.7649492 , -0.66997206, -0.42259073,\n",
      "        -0.75030595, -0.5528497 , -0.653441  , -0.5219667 , -0.55036193,\n",
      "        -0.4115535 , -0.5915975 , -0.5096015 , -0.5566663 , -0.39612812,\n",
      "        -0.34296218, -0.4913441 , -0.50283307, -0.48371625, -0.4034294 ,\n",
      "        -0.6975335 , -0.7425752 , -0.49024788, -0.542024  , -0.2848799 ,\n",
      "        -0.5804699 , -0.43209678, -0.5135337 , -0.47872555, -0.60387194,\n",
      "        -0.4938911 , -0.38207942, -0.48180544, -0.5230396 , -0.4158064 ,\n",
      "        -0.6591399 , -0.37792954, -0.4116072 , -0.55662537, -0.5867378 ,\n",
      "        -0.39524287, -0.43465582, -0.49194676, -0.37556377, -0.38926896,\n",
      "        -0.40040532, -0.5666718 , -0.62566084, -0.40276706, -0.48160064,\n",
      "        -0.55837953, -0.72155875, -0.405821  , -0.54017466, -0.7622776 ,\n",
      "        -0.5904732 , -0.5057772 , -0.6820936 , -0.47650158, -0.4600595 ,\n",
      "        -0.6608286 , -0.652446  , -0.4950025 , -0.40435955, -0.74502337,\n",
      "        -0.5246923 , -0.6330568 , -0.5295229 , -0.5590599 , -0.33067435,\n",
      "        -0.5028657 , -0.49822995, -0.72676545, -0.5765192 , -0.40687418,\n",
      "        -0.41769052, -0.5470686 , -0.5118857 , -0.39757925, -0.69446063,\n",
      "        -0.36933225, -0.47475824, -0.50806254, -0.489954  , -0.28150353,\n",
      "        -0.54311407, -0.40688443, -0.48005164],\n",
      "       [-0.3931989 , -0.61554027, -0.44959924, -0.3537691 , -0.47846445,\n",
      "        -0.5989333 , -0.46940225, -0.37279308, -0.43572924, -0.29370284,\n",
      "        -0.61839163, -0.38949022, -0.49376604, -0.43721002, -0.52413344,\n",
      "        -0.30360043, -0.48908997, -0.4789895 , -0.39121875, -0.48145425,\n",
      "        -0.40343595, -0.2882032 , -0.45271662, -0.50405914, -0.45083812,\n",
      "        -0.3920733 , -0.39626583, -0.36567956, -0.47410554, -0.56916595,\n",
      "        -0.35524294, -0.52508456, -0.5690911 , -0.39827737, -0.499216  ,\n",
      "        -0.4493328 , -0.47852486, -0.52317417, -0.5433292 , -0.49969473,\n",
      "        -0.42787883, -0.30844808, -0.752368  , -0.656608  , -0.2944803 ,\n",
      "        -0.68581706, -0.48736262, -0.46291986, -0.49280283, -0.47553295,\n",
      "        -0.41928113, -0.43463755, -0.39038754, -0.53816795, -0.37942818,\n",
      "        -0.449807  , -0.5173886 , -0.5102746 , -0.55173886, -0.45779628,\n",
      "        -0.48528492, -0.5687011 , -0.41372955, -0.30971298, -0.42185137,\n",
      "        -0.52699214, -0.36948642, -0.432338  , -0.4070984 , -0.39783186,\n",
      "        -0.44983572, -0.3971023 , -0.394549  , -0.57082623, -0.4226523 ,\n",
      "        -0.43644342, -0.3260424 , -0.38800532, -0.52649266, -0.5739197 ,\n",
      "        -0.43127564, -0.2765774 , -0.47330415, -0.38306427, -0.44792578,\n",
      "        -0.28741953, -0.3763738 , -0.45024088, -0.5275883 , -0.3584142 ,\n",
      "        -0.43202883, -0.5801787 , -0.50997615, -0.44326842, -0.6512312 ,\n",
      "        -0.48445883, -0.34746718, -0.50435215, -0.4917546 , -0.36923745,\n",
      "        -0.41262063, -0.49772578, -0.36986163, -0.3615504 , -0.49585858,\n",
      "        -0.48216075, -0.4894863 , -0.31533733, -0.5140495 , -0.34360504,\n",
      "        -0.42141545, -0.3494848 , -0.51435906, -0.48705977, -0.33211318,\n",
      "        -0.26440567, -0.48557255, -0.43040705, -0.52354926, -0.46492746,\n",
      "        -0.37126938, -0.4576324 , -0.42857957, -0.39003518, -0.4154784 ,\n",
      "        -0.5474794 , -0.5125282 , -0.416343  ],\n",
      "       [-0.56011957, -0.68436986, -0.44236085, -0.53683704, -0.4627178 ,\n",
      "        -0.7414797 , -0.5308222 , -0.45897317, -0.41143557, -0.20796502,\n",
      "        -0.49499568, -0.3409917 , -0.50400174, -0.57132417, -0.5595612 ,\n",
      "        -0.6169778 , -0.3747416 , -0.42241928, -0.39334047, -0.51488674,\n",
      "        -0.46762618, -0.53410393, -0.55896354, -0.5712803 , -0.49909297,\n",
      "        -0.4067452 , -0.32957548, -0.32923537, -0.5314926 , -0.6583323 ,\n",
      "        -0.51932424, -0.5777176 , -0.47660044, -0.36356434, -0.78551066,\n",
      "        -0.46695635, -0.6642079 , -0.5969971 , -0.58571714, -0.5930848 ,\n",
      "        -0.45639145, -0.36392406, -0.89500743, -0.6275469 , -0.43864006,\n",
      "        -0.5533429 , -0.6859828 , -0.65150726, -0.5985649 , -0.6061236 ,\n",
      "        -0.51278865, -0.52390975, -0.38131315, -0.39360246, -0.44927847,\n",
      "        -0.51378155, -0.34751463, -0.6377363 , -0.6333994 , -0.5936411 ,\n",
      "        -0.50906414, -0.9552382 , -0.5584466 , -0.45224103, -0.38911656,\n",
      "        -0.31874463, -0.5111166 , -0.47726664, -0.4273844 , -0.6014075 ,\n",
      "        -0.48532343, -0.41274974, -0.5307681 , -0.54818845, -0.46970174,\n",
      "        -0.5900764 , -0.36793116, -0.38171428, -0.6009016 , -0.49162942,\n",
      "        -0.50307655, -0.35663617, -0.556164  , -0.2912777 , -0.47535136,\n",
      "        -0.500977  , -0.47526768, -0.5856898 , -0.574118  , -0.4808902 ,\n",
      "        -0.47050112, -0.6863338 , -0.49292773, -0.47290903, -0.5512863 ,\n",
      "        -0.50527585, -0.4050639 , -0.49153125, -0.45637435, -0.3104    ,\n",
      "        -0.58940524, -0.6421088 , -0.43425724, -0.37771446, -0.8915428 ,\n",
      "        -0.4552511 , -0.6083358 , -0.48852468, -0.47641453, -0.45930517,\n",
      "        -0.34197864, -0.36829594, -0.797902  , -0.49758196, -0.42131096,\n",
      "        -0.38368317, -0.651073  , -0.4877744 , -0.6197328 , -0.7817297 ,\n",
      "        -0.37214494, -0.3412182 , -0.4390648 , -0.527085  , -0.60511047,\n",
      "        -0.60699356, -0.17871769, -0.59978676]], dtype=float32), 'preds': [['天', '亮', '以', '前', '说', '再', '见', '，', '笑', '着', '泪', '流', '满', '面', '，', '去', '迎', '接', '应', '该', '你', '的', '，', '更', '好', '的', '明', '天', '[SEP]'], ['昙', '花', '若', '只', '一', '现', '，', '更', '要', '开', '的', '耀', '眼', '，', '别', '回', '头', '去', '拥', '有', '，', '属', '于', '你', '更', '好', '的', '世', '界', '[SEP]'], ['天', '亮', '以', '前', '说', '再', '见', '，', '让', '我', '留', '在', '今', '天', '，', '去', '保', '护', '我', '和', '你', '的', '，', '最', '好', '的', '夏', '天', '[SEP]'], ['情', '丝', '若', '水', '三', '千', '，', '只', '取', '一', '瓢', '眷', '恋', '，', '当', '你', '来', '过', '的', '纪', '念', '[SEP]']]}\n"
     ]
    }
   ],
   "source": [
    "print(model.predict(X))    # 充分收敛后，输入和输出将基本一致"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fundamental-poverty",
   "metadata": {},
   "source": [
    "推理时可根据需要调整噪声大小，生成错误文本，提供给其他NLP任务"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "accessory-lesbian",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Running inference on 4 samples\n",
      "INFO:tensorflow:process 100.0%, 23.50 examples/sec\n",
      "{'miu': array([[-0.4396022 ,  0.53117496,  0.17832753,  0.3916682 ,  0.5434318 ,\n",
      "        -0.349801  ,  0.4876728 ,  0.06779953,  0.16947743, -0.2487141 ,\n",
      "        -0.47011247, -0.35064802,  0.38101462, -0.39234826,  0.76904225,\n",
      "         0.7980169 , -0.31436703, -0.6142652 , -0.33549577, -0.33714166,\n",
      "         0.73015195,  0.47610235,  0.00210417, -0.09950063,  0.18604045,\n",
      "        -0.11020187, -0.48238188,  0.14054313, -0.51977366, -0.6530189 ,\n",
      "         0.29201305, -0.11135881, -0.6522274 , -0.01328439, -0.6988805 ,\n",
      "        -0.1957497 , -0.61858803,  0.4604739 , -0.16431828, -0.64519733,\n",
      "        -0.28615648, -0.10812853,  0.8175854 , -0.15867183,  0.23407696,\n",
      "         0.25935042,  0.48445275,  0.7359865 , -0.4633496 ,  0.03561433,\n",
      "        -0.00894808,  0.34221044,  0.46491835, -0.5450153 ,  0.3270287 ,\n",
      "        -0.4716175 , -0.03313906,  0.11883581, -0.28118053,  0.53394926,\n",
      "         0.47560224, -0.7123323 , -0.28942183,  0.01549871, -0.6245871 ,\n",
      "         0.11586095,  0.3069579 , -0.09454164,  0.24998963, -0.49753317,\n",
      "        -0.36490378, -0.0467333 ,  0.16291521,  0.03164184, -0.02466292,\n",
      "         0.77220887,  0.08216386, -0.39379105,  0.22965193, -0.6106302 ,\n",
      "         0.12905608,  0.03143375,  0.2750468 , -0.11824985,  0.11386785,\n",
      "        -0.21679497, -0.30680424,  0.15367134, -0.23380323, -0.645449  ,\n",
      "         0.54022264,  0.7515234 ,  0.17085446, -0.32974073, -0.5721658 ,\n",
      "         0.45940095, -0.2601403 ,  0.08977263,  0.01357785,  0.28882757,\n",
      "         0.02907572, -0.54030764,  0.23230802, -0.16057894,  0.8285682 ,\n",
      "        -0.21954456, -0.23607539, -0.3354532 ,  0.6094573 ,  0.01860696,\n",
      "        -0.3252273 , -0.06578828, -0.82433325,  0.16053909,  0.11647375,\n",
      "        -0.4680446 ,  0.41359082,  0.10610711, -0.05752654, -0.82179433,\n",
      "        -0.4556519 ,  0.15392034,  0.40726206,  0.22530027, -0.2509305 ,\n",
      "         0.05658961, -0.01185258, -0.3808937 ],\n",
      "       [-0.40424427, -0.6613178 , -0.55455583, -0.54453754,  0.15113   ,\n",
      "        -0.7264001 , -0.1964842 , -0.12757275,  0.3063119 , -0.09371693,\n",
      "        -0.78356874, -0.177179  , -0.61630243,  0.57635134,  0.02743418,\n",
      "        -0.6013617 ,  0.7516876 , -0.51784784, -0.62055796, -0.40939158,\n",
      "        -0.17829244,  0.13516863, -0.08669522,  0.4725463 ,  0.00599788,\n",
      "        -0.16556753, -0.4814041 , -0.09986823,  0.3095039 , -0.8197245 ,\n",
      "         0.35185435, -0.76147985,  0.22937186,  0.18708481,  0.29995713,\n",
      "         0.15827122,  0.02635231,  0.27391377,  0.5452139 ,  0.3958741 ,\n",
      "        -0.3562757 ,  0.0354724 ,  0.05675625, -0.4242439 , -0.02828944,\n",
      "        -0.56530756,  0.16176453, -0.31689498,  0.01594808,  0.47264296,\n",
      "         0.57770824,  0.14317366, -0.20788841,  0.12255213, -0.34458688,\n",
      "        -0.05454639,  0.61516744, -0.67166615,  0.7109099 ,  0.06450709,\n",
      "         0.13677163, -0.25293982, -0.4495866 ,  0.39292803, -0.08624236,\n",
      "        -0.42750838,  0.40305406,  0.5100008 , -0.5650745 , -0.39525053,\n",
      "        -0.29834765,  0.32661283,  0.38759992, -0.4559074 ,  0.75325495,\n",
      "         0.31332484, -0.07387909,  0.52348846,  0.60346425,  0.25267062,\n",
      "         0.59518296, -0.06666263,  0.14753018, -0.1554321 ,  0.39860487,\n",
      "        -0.5016212 , -0.20139454, -0.6015193 ,  0.8241274 , -0.4132846 ,\n",
      "         0.22998306,  0.4072479 , -0.7947555 , -0.2875811 , -0.49485567,\n",
      "        -0.07750013,  0.3428134 ,  0.50484043,  0.70465666, -0.5235935 ,\n",
      "         0.22626197,  0.16179188,  0.12681942, -0.2762881 ,  0.8757036 ,\n",
      "         0.6251476 , -0.34778562, -0.5337127 , -0.23631097, -0.07493189,\n",
      "         0.13019995,  0.37212723, -0.15985766,  0.35012403, -0.17318068,\n",
      "        -0.04319626,  0.6262446 , -0.7358355 ,  0.8688896 , -0.6850959 ,\n",
      "         0.15048127,  0.56519634,  0.3139976 , -0.7050873 , -0.54715115,\n",
      "         0.77197987, -0.4063307 , -0.56046206],\n",
      "       [-0.3642728 ,  0.46004498,  0.01059638,  0.32406574,  0.53835565,\n",
      "        -0.35064748,  0.45198888,  0.07451502,  0.01790816, -0.21773584,\n",
      "        -0.4787502 , -0.38566804,  0.17716734, -0.23125465,  0.6740272 ,\n",
      "         0.7275375 , -0.21396048, -0.6463626 , -0.48730168, -0.48037007,\n",
      "         0.6404515 ,  0.36057398, -0.01316697,  0.09231912,  0.29894668,\n",
      "        -0.0330594 , -0.6024469 ,  0.07630484, -0.38744673, -0.667308  ,\n",
      "         0.26080528, -0.1799443 , -0.50601375, -0.03865441, -0.58848375,\n",
      "        -0.178581  , -0.52887404,  0.36266178,  0.05854804, -0.51894945,\n",
      "        -0.35518825,  0.00790745,  0.72956234, -0.29058093,  0.0846502 ,\n",
      "         0.08107164,  0.5253183 ,  0.65459776, -0.3534064 ,  0.13854381,\n",
      "        -0.03686604,  0.39595297,  0.47555503, -0.35170725,  0.18984316,\n",
      "        -0.3834523 ,  0.10976946,  0.07115378, -0.12964062,  0.49872229,\n",
      "         0.5209197 , -0.5893627 , -0.38104895,  0.07993336, -0.52138287,\n",
      "        -0.10246953,  0.35107628, -0.09562219,  0.20706494, -0.41679892,\n",
      "        -0.45118335,  0.07654575,  0.05550493, -0.05588711,  0.16146721,\n",
      "         0.6882174 , -0.06010684, -0.30950406,  0.36571613, -0.510865  ,\n",
      "         0.03544237, -0.03008462,  0.36853167, -0.06452294,  0.15827432,\n",
      "        -0.23604253, -0.2281504 , -0.0185689 , -0.20900336, -0.57493144,\n",
      "         0.427282  ,  0.7825432 ,  0.07771046, -0.429275  , -0.64938486,\n",
      "         0.2931583 , -0.2670598 , -0.08133707,  0.05289136,  0.06665377,\n",
      "        -0.06810778, -0.51975286,  0.24316153, -0.04403505,  0.84778327,\n",
      "        -0.16769047, -0.15489088, -0.38689238,  0.51435137,  0.1327065 ,\n",
      "        -0.38551414, -0.11884423, -0.77674544,  0.1371838 ,  0.02346713,\n",
      "        -0.43203786,  0.49232924, -0.08234139, -0.02754244, -0.8178904 ,\n",
      "        -0.4164275 ,  0.32128194,  0.35656524,  0.21968696, -0.22027534,\n",
      "         0.101785  , -0.23617807, -0.3583702 ],\n",
      "       [ 0.51699   , -0.05625712, -0.46980837,  0.1681818 ,  0.2717811 ,\n",
      "        -0.24365702,  0.23892926, -0.4320021 , -0.7521961 ,  0.05039456,\n",
      "        -0.58603185, -0.19359787, -0.52544624,  0.3090927 , -0.5597222 ,\n",
      "        -0.10307791,  0.01066909, -0.4828461 , -0.5633853 , -0.7958153 ,\n",
      "        -0.50360787, -0.25945395, -0.36204723,  0.7365016 ,  0.74419653,\n",
      "         0.33362523, -0.69249564, -0.47149196,  0.53172225, -0.5156852 ,\n",
      "        -0.26261383,  0.04042988,  0.71059215, -0.18021889,  0.5622784 ,\n",
      "         0.08040801,  0.41929275, -0.48548552,  0.5546837 ,  0.484002  ,\n",
      "        -0.10581797,  0.19771634, -0.6374899 , -0.8068022 , -0.6890977 ,\n",
      "        -0.63238823,  0.54130673, -0.17360899,  0.43616864,  0.4141887 ,\n",
      "        -0.5260514 ,  0.53272367,  0.4639936 ,  0.77233076, -0.50562793,\n",
      "         0.14527456,  0.52342486,  0.3969197 ,  0.3626948 , -0.04184609,\n",
      "         0.5647089 ,  0.72535104, -0.5060267 ,  0.13688307,  0.4256356 ,\n",
      "        -0.64741695, -0.05520255, -0.33446285,  0.0542817 ,  0.484301  ,\n",
      "        -0.62048656,  0.3161051 , -0.7106597 , -0.40326267,  0.6463302 ,\n",
      "        -0.29725516, -0.50547385,  0.25041732,  0.5518853 ,  0.69279885,\n",
      "        -0.6185073 , -0.47554046,  0.6717064 ,  0.25623134,  0.2763538 ,\n",
      "        -0.08276364, -0.32833782, -0.6685877 , -0.52612495,  0.01979494,\n",
      "        -0.44114685,  0.57690847, -0.19277796, -0.6443098 , -0.65701175,\n",
      "        -0.7466271 , -0.0226044 , -0.73293847, -0.29825896, -0.4849551 ,\n",
      "        -0.4149438 , -0.3612215 ,  0.3061612 ,  0.26462027,  0.6999105 ,\n",
      "         0.03967398,  0.5756744 , -0.31545925,  0.03272838,  0.51458156,\n",
      "        -0.6775815 , -0.19910872,  0.298664  , -0.02839263, -0.24655114,\n",
      "        -0.43876162,  0.33000016, -0.5196884 , -0.20472391, -0.03096777,\n",
      "        -0.22015893,  0.5531062 , -0.4260221 ,  0.46003515,  0.23697607,\n",
      "        -0.20001571, -0.780906  ,  0.08472728]], dtype=float32), 'sigma': array([[-0.3831867 , -0.52395594, -0.39490977, -0.33780122, -0.41583732,\n",
      "        -0.47149825, -0.40949988, -0.34398192, -0.32951543, -0.31029698,\n",
      "        -0.54492164, -0.41599876, -0.4324875 , -0.3431378 , -0.45812592,\n",
      "        -0.25838342, -0.45791572, -0.42541614, -0.40994573, -0.48496142,\n",
      "        -0.37105954, -0.2794265 , -0.37017506, -0.55627453, -0.35677972,\n",
      "        -0.37343624, -0.383798  , -0.38137573, -0.40318215, -0.5163089 ,\n",
      "        -0.32750708, -0.42829078, -0.5289174 , -0.33509782, -0.49119   ,\n",
      "        -0.38223416, -0.41587228, -0.46118608, -0.52699506, -0.4486439 ,\n",
      "        -0.35629532, -0.33426616, -0.68044096, -0.5730364 , -0.2363012 ,\n",
      "        -0.63126826, -0.45245185, -0.4358171 , -0.44235292, -0.43667004,\n",
      "        -0.4293162 , -0.35206518, -0.3358361 , -0.5113451 , -0.2972256 ,\n",
      "        -0.36040854, -0.52424973, -0.49219388, -0.43922335, -0.32743767,\n",
      "        -0.40915126, -0.4421557 , -0.3465399 , -0.24146679, -0.3643712 ,\n",
      "        -0.4596647 , -0.3304459 , -0.4692658 , -0.29292592, -0.3044884 ,\n",
      "        -0.37971488, -0.38895005, -0.3393326 , -0.47503147, -0.3558076 ,\n",
      "        -0.3351344 , -0.2907724 , -0.31920582, -0.42375913, -0.5482773 ,\n",
      "        -0.3652358 , -0.24286714, -0.44433093, -0.39158028, -0.36590925,\n",
      "        -0.29350862, -0.34788963, -0.38077998, -0.47669187, -0.3440472 ,\n",
      "        -0.36944002, -0.52661586, -0.44760892, -0.4727469 , -0.54288644,\n",
      "        -0.42217866, -0.2958092 , -0.41930237, -0.507913  , -0.39832836,\n",
      "        -0.33301958, -0.48628405, -0.31339568, -0.3857982 , -0.40353784,\n",
      "        -0.3976823 , -0.4355631 , -0.25579527, -0.47944912, -0.33822495,\n",
      "        -0.42450094, -0.29456958, -0.4118774 , -0.43760365, -0.29186782,\n",
      "        -0.19793081, -0.43110508, -0.4421649 , -0.45649183, -0.30406314,\n",
      "        -0.32333088, -0.40880707, -0.38601798, -0.3375841 , -0.35035115,\n",
      "        -0.47714743, -0.45460117, -0.30579388],\n",
      "       [-0.45851618, -0.8126475 , -0.5778383 , -0.4340984 , -0.38926283,\n",
      "        -0.6052165 , -0.504553  , -0.48874742, -0.44569477, -0.3518301 ,\n",
      "        -0.58542645, -0.42095643, -0.6020549 , -0.62612903, -0.54076815,\n",
      "        -0.6840471 , -0.41007262, -0.6015291 , -0.47547355, -0.34628716,\n",
      "        -0.5558731 , -0.38952506, -0.45283544, -0.53094584, -0.41673958,\n",
      "        -0.5672391 , -0.5418804 , -0.3373803 , -0.6062407 , -0.5406409 ,\n",
      "        -0.3945115 , -0.46906376, -0.6198423 , -0.40681604, -0.5748803 ,\n",
      "        -0.5177635 , -0.5943926 , -0.611834  , -0.6578959 , -0.73868144,\n",
      "        -0.3899627 , -0.3567461 , -0.7649492 , -0.66997206, -0.42259073,\n",
      "        -0.75030595, -0.5528497 , -0.653441  , -0.5219667 , -0.55036193,\n",
      "        -0.4115535 , -0.5915975 , -0.5096015 , -0.5566663 , -0.39612812,\n",
      "        -0.34296218, -0.4913441 , -0.50283307, -0.48371625, -0.4034294 ,\n",
      "        -0.6975335 , -0.7425752 , -0.49024788, -0.542024  , -0.2848799 ,\n",
      "        -0.5804699 , -0.43209678, -0.5135337 , -0.47872555, -0.60387194,\n",
      "        -0.4938911 , -0.38207942, -0.48180544, -0.5230396 , -0.4158064 ,\n",
      "        -0.6591399 , -0.37792954, -0.4116072 , -0.55662537, -0.5867378 ,\n",
      "        -0.39524287, -0.43465582, -0.49194676, -0.37556377, -0.38926896,\n",
      "        -0.40040532, -0.5666718 , -0.62566084, -0.40276706, -0.48160064,\n",
      "        -0.55837953, -0.72155875, -0.405821  , -0.54017466, -0.7622776 ,\n",
      "        -0.5904732 , -0.5057772 , -0.6820936 , -0.47650158, -0.4600595 ,\n",
      "        -0.6608286 , -0.652446  , -0.4950025 , -0.40435955, -0.74502337,\n",
      "        -0.5246923 , -0.6330568 , -0.5295229 , -0.5590599 , -0.33067435,\n",
      "        -0.5028657 , -0.49822995, -0.72676545, -0.5765192 , -0.40687418,\n",
      "        -0.41769052, -0.5470686 , -0.5118857 , -0.39757925, -0.69446063,\n",
      "        -0.36933225, -0.47475824, -0.50806254, -0.489954  , -0.28150353,\n",
      "        -0.54311407, -0.40688443, -0.48005164],\n",
      "       [-0.3931989 , -0.61554027, -0.44959924, -0.3537691 , -0.47846445,\n",
      "        -0.5989333 , -0.46940225, -0.37279308, -0.43572924, -0.29370284,\n",
      "        -0.61839163, -0.38949022, -0.49376604, -0.43721002, -0.52413344,\n",
      "        -0.30360043, -0.48908997, -0.4789895 , -0.39121875, -0.48145425,\n",
      "        -0.40343595, -0.2882032 , -0.45271662, -0.50405914, -0.45083812,\n",
      "        -0.3920733 , -0.39626583, -0.36567956, -0.47410554, -0.56916595,\n",
      "        -0.35524294, -0.52508456, -0.5690911 , -0.39827737, -0.499216  ,\n",
      "        -0.4493328 , -0.47852486, -0.52317417, -0.5433292 , -0.49969473,\n",
      "        -0.42787883, -0.30844808, -0.752368  , -0.656608  , -0.2944803 ,\n",
      "        -0.68581706, -0.48736262, -0.46291986, -0.49280283, -0.47553295,\n",
      "        -0.41928113, -0.43463755, -0.39038754, -0.53816795, -0.37942818,\n",
      "        -0.449807  , -0.5173886 , -0.5102746 , -0.55173886, -0.45779628,\n",
      "        -0.48528492, -0.5687011 , -0.41372955, -0.30971298, -0.42185137,\n",
      "        -0.52699214, -0.36948642, -0.432338  , -0.4070984 , -0.39783186,\n",
      "        -0.44983572, -0.3971023 , -0.394549  , -0.57082623, -0.4226523 ,\n",
      "        -0.43644342, -0.3260424 , -0.38800532, -0.52649266, -0.5739197 ,\n",
      "        -0.43127564, -0.2765774 , -0.47330415, -0.38306427, -0.44792578,\n",
      "        -0.28741953, -0.3763738 , -0.45024088, -0.5275883 , -0.3584142 ,\n",
      "        -0.43202883, -0.5801787 , -0.50997615, -0.44326842, -0.6512312 ,\n",
      "        -0.48445883, -0.34746718, -0.50435215, -0.4917546 , -0.36923745,\n",
      "        -0.41262063, -0.49772578, -0.36986163, -0.3615504 , -0.49585858,\n",
      "        -0.48216075, -0.4894863 , -0.31533733, -0.5140495 , -0.34360504,\n",
      "        -0.42141545, -0.3494848 , -0.51435906, -0.48705977, -0.33211318,\n",
      "        -0.26440567, -0.48557255, -0.43040705, -0.52354926, -0.46492746,\n",
      "        -0.37126938, -0.4576324 , -0.42857957, -0.39003518, -0.4154784 ,\n",
      "        -0.5474794 , -0.5125282 , -0.416343  ],\n",
      "       [-0.56011957, -0.68436986, -0.44236085, -0.53683704, -0.4627178 ,\n",
      "        -0.7414797 , -0.5308222 , -0.45897317, -0.41143557, -0.20796502,\n",
      "        -0.49499568, -0.3409917 , -0.50400174, -0.57132417, -0.5595612 ,\n",
      "        -0.6169778 , -0.3747416 , -0.42241928, -0.39334047, -0.51488674,\n",
      "        -0.46762618, -0.53410393, -0.55896354, -0.5712803 , -0.49909297,\n",
      "        -0.4067452 , -0.32957548, -0.32923537, -0.5314926 , -0.6583323 ,\n",
      "        -0.51932424, -0.5777176 , -0.47660044, -0.36356434, -0.78551066,\n",
      "        -0.46695635, -0.6642079 , -0.5969971 , -0.58571714, -0.5930848 ,\n",
      "        -0.45639145, -0.36392406, -0.89500743, -0.6275469 , -0.43864006,\n",
      "        -0.5533429 , -0.6859828 , -0.65150726, -0.5985649 , -0.6061236 ,\n",
      "        -0.51278865, -0.52390975, -0.38131315, -0.39360246, -0.44927847,\n",
      "        -0.51378155, -0.34751463, -0.6377363 , -0.6333994 , -0.5936411 ,\n",
      "        -0.50906414, -0.9552382 , -0.5584466 , -0.45224103, -0.38911656,\n",
      "        -0.31874463, -0.5111166 , -0.47726664, -0.4273844 , -0.6014075 ,\n",
      "        -0.48532343, -0.41274974, -0.5307681 , -0.54818845, -0.46970174,\n",
      "        -0.5900764 , -0.36793116, -0.38171428, -0.6009016 , -0.49162942,\n",
      "        -0.50307655, -0.35663617, -0.556164  , -0.2912777 , -0.47535136,\n",
      "        -0.500977  , -0.47526768, -0.5856898 , -0.574118  , -0.4808902 ,\n",
      "        -0.47050112, -0.6863338 , -0.49292773, -0.47290903, -0.5512863 ,\n",
      "        -0.50527585, -0.4050639 , -0.49153125, -0.45637435, -0.3104    ,\n",
      "        -0.58940524, -0.6421088 , -0.43425724, -0.37771446, -0.8915428 ,\n",
      "        -0.4552511 , -0.6083358 , -0.48852468, -0.47641453, -0.45930517,\n",
      "        -0.34197864, -0.36829594, -0.797902  , -0.49758196, -0.42131096,\n",
      "        -0.38368317, -0.651073  , -0.4877744 , -0.6197328 , -0.7817297 ,\n",
      "        -0.37214494, -0.3412182 , -0.4390648 , -0.527085  , -0.60511047,\n",
      "        -0.60699356, -0.17871769, -0.59978676]], dtype=float32), 'preds': [['天', '亮', '以', '瓢', '说', '再', '见', '，', '让', '我', '泪', '在', '满', '面', '，', '去', '保', '护', '应', '和', '你', '的', '，', '更', '好', '的', '明', '天', '[SEP]'], ['昙', '花', '若', '念', '一', '现', '，', '更', '要', '开', '的', '耀', '满', '，', '别', '去', '恋', '去', '拥', '有', '，', '属', '于', '你', '更', '好', '的', '世', '界', '[SEP]'], ['天', '亮', '以', '前', '说', '再', '见', '，', '让', '我', '留', '流', '今', '天', '，', '去', '迎', '护', '我', '该', '你', '的', '，', '更', '好', '的', '明', '天', '[SEP]'], ['取', '丝', '以', '水', '三', '千', '，', '只', '亮', '一', '瓢', '眷', '恋', '，', '当', '你', '来', '该', '的', '纪', '念', '[SEP]']]}\n"
     ]
    }
   ],
   "source": [
    "print(model.predict(X, bias=3))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.13 64-bit",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  },
  "vscode": {
   "interpreter": {
    "hash": "265fd6f62f200408acbbeae0248f34bed9f93569a643842b7a25d2cd76cae5e5"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}