 View on TensorFlow.org
|
 Run in Google Colab
|
 View on GitHub
|
 Download notebook
|
 查看 TF Hub 模型 |
TF-Hub 是一个平台，用于共享打包为可重用资源的机器学习专业知识，尤其是经过预训练的模块。在本教程中，我们将使用 TF-Hub 文本嵌入向量模块来训练具有合理基线准确率的简单情感分类器。之后，我们会将预测结果提交给 Kaggle。
有关使用 TF-Hub 进行文本分类的更详细教程，以及提高准确率的后续步骤，请查看使用 TF-Hub 进行分本分类。
设置
pip install -q kaggle
import tensorflow as tf
import tensorflow_hub as hub
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
import zipfile
from sklearn import model_selection
由于本教程将使用 Kaggle 中的数据集，因此需要为您的 Kaggle 帐号创建 API 令牌，并将其上传到 Colab 环境。
import os
import pathlib
# Upload the API token.
def get_kaggle():
try:
import kaggle
return kaggle
except OSError:
pass
token_file = pathlib.Path("~/.kaggle/kaggle.json").expanduser()
token_file.parent.mkdir(exist_ok=True, parents=True)
try:
from google.colab import files
except ImportError:
raise ValueError("Could not find kaggle token.")
uploaded = files.upload()
token_content = uploaded.get('kaggle.json', None)
if token_content:
token_file.write_bytes(token_content)
token_file.chmod(0o600)
else:
raise ValueError('Need a file named "kaggle.json"')
import kaggle
return kaggle
kaggle = get_kaggle()
开始
数据
我们将尝试完成 Kaggle 的 Sentiment Analysis on Movie Reviews 任务。数据集由 Rotten Tomatoes 电影评论中符合句法的子短语组成。任务是采用五分制将短语标记为负面或正面。
在使用此 API 下载数据之前，您必须接受竞赛规则。
SENTIMENT_LABELS = [
"negative", "somewhat negative", "neutral", "somewhat positive", "positive"
]
# Add a column with readable values representing the sentiment.
def add_readable_labels_column(df, sentiment_value_column):
df["SentimentLabel"] = df[sentiment_value_column].replace(
range(5), SENTIMENT_LABELS)
# Download data from Kaggle and create a DataFrame.
def load_data_from_zip(path):
with zipfile.ZipFile(path, "r") as zip_ref:
name = zip_ref.namelist()[0]
with zip_ref.open(name) as zf:
return pd.read_csv(zf, sep="\t", index_col=0)
# The data does not come with a validation set so we'll create one from the
# training set.
def get_data(competition, train_file, test_file, validation_set_ratio=0.1):
data_path = pathlib.Path("data")
kaggle.api.competition_download_files(competition, data_path)
competition_path = (data_path/competition)
competition_path.mkdir(exist_ok=True, parents=True)
competition_zip_path = competition_path.with_suffix(".zip")
with zipfile.ZipFile(competition_zip_path, "r") as zip_ref:
zip_ref.extractall(competition_path)
train_df = load_data_from_zip(competition_path/train_file)
test_df = load_data_from_zip(competition_path/test_file)
# Add a human readable label.
add_readable_labels_column(train_df, "Sentiment")
# We split by sentence ids, because we don't want to have phrases belonging
# to the same sentence in both training and validation set.
train_indices, validation_indices = model_selection.train_test_split(
np.unique(train_df["SentenceId"]),
test_size=validation_set_ratio,
random_state=0)
validation_df = train_df[train_df["SentenceId"].isin(validation_indices)]
train_df = train_df[train_df["SentenceId"].isin(train_indices)]
print("Split the training data into %d training and %d validation examples." %
(len(train_df), len(validation_df)))
return train_df, validation_df, test_df
train_df, validation_df, test_df = get_data(
"sentiment-analysis-on-movie-reviews",
"train.tsv.zip", "test.tsv.zip")
Split the training data into 140315 training and 15745 validation examples.
注：本竞赛的任务不是对整个评论进行评分，而是对评论中的各个短语进行评分。这是一项更加艰巨的任务。
train_df.head(20)
训练模型
注：我们也可以将此任务建模为回归模型，请参阅使用 TF-Hub 进行文本分类。
class MyModel(tf.keras.Model):
def __init__(self, hub_url):
super().__init__()
self.hub_url = hub_url
self.embed = hub.load(self.hub_url).signatures['default']
self.sequential = tf.keras.Sequential([
tf.keras.layers.Dense(500),
tf.keras.layers.Dense(100),
tf.keras.layers.Dense(5),
])
def call(self, inputs):
phrases = inputs['Phrase'][:,0]
embedding = 5*self.embed(phrases)['default']
return self.sequential(embedding)
def get_config(self):
return {"hub_url":self.hub_url}
model = MyModel("https://tfhub.dev/google/nnlm-en-dim128/1")
model.compile(
loss = tf.losses.SparseCategoricalCrossentropy(from_logits=True),
optimizer=tf.optimizers.Adam(),
metrics = [tf.keras.metrics.SparseCategoricalAccuracy(name="accuracy")])
2021-08-13 20:43:34.606466: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:34.613478: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:34.614419: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:34.616244: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 AVX512F FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags. 2021-08-13 20:43:34.616765: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:34.617726: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:34.618627: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:35.193388: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:35.194452: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:35.195419: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero 2021-08-13 20:43:35.196250: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1510] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 14648 MB memory: -> device: 0, name: Tesla V100-SXM2-16GB, pci bus id: 0000:00:05.0, compute capability: 7.0 2021-08-13 20:43:35.523279: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:185] None of the MLIR Optimization Passes are enabled (registered 2)
history = model.fit(x=dict(train_df), y=train_df['Sentiment'],
validation_data=(dict(validation_df), validation_df['Sentiment']),
epochs = 25)
Epoch 1/25 4385/4385 [==============================] - 14s 3ms/step - loss: 1.0245 - accuracy: 0.5855 - val_loss: 0.9837 - val_accuracy: 0.5942 Epoch 2/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9998 - accuracy: 0.5948 - val_loss: 0.9969 - val_accuracy: 0.5885 Epoch 3/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9950 - accuracy: 0.5968 - val_loss: 0.9856 - val_accuracy: 0.5957 Epoch 4/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9926 - accuracy: 0.5977 - val_loss: 0.9857 - val_accuracy: 0.5973 Epoch 5/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9916 - accuracy: 0.5983 - val_loss: 0.9945 - val_accuracy: 0.5891 Epoch 6/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9907 - accuracy: 0.5976 - val_loss: 0.9822 - val_accuracy: 0.5932 Epoch 7/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9899 - accuracy: 0.5973 - val_loss: 0.9793 - val_accuracy: 0.5907 Epoch 8/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9889 - accuracy: 0.5993 - val_loss: 0.9811 - val_accuracy: 0.5952 Epoch 9/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9888 - accuracy: 0.5984 - val_loss: 0.9805 - val_accuracy: 0.5915 Epoch 10/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9883 - accuracy: 0.5985 - val_loss: 0.9824 - val_accuracy: 0.5985 Epoch 11/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9877 - accuracy: 0.5984 - val_loss: 0.9760 - val_accuracy: 0.5985 Epoch 12/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9877 - accuracy: 0.5990 - val_loss: 0.9824 - val_accuracy: 0.5967 Epoch 13/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9877 - accuracy: 0.5990 - val_loss: 0.9781 - val_accuracy: 0.5971 Epoch 14/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9870 - accuracy: 0.5993 - val_loss: 0.9817 - val_accuracy: 0.5975 Epoch 15/25 4385/4385 [==============================] - 12s 3ms/step - loss: 0.9869 - accuracy: 0.5990 - val_loss: 0.9792 - val_accuracy: 0.5986 Epoch 16/25 4385/4385 [==============================] - 12s 3ms/step - loss: 0.9872 - accuracy: 0.6000 - val_loss: 0.9790 - val_accuracy: 0.5980 Epoch 17/25 4385/4385 [==============================] - 12s 3ms/step - loss: 0.9866 - accuracy: 0.5991 - val_loss: 0.9765 - val_accuracy: 0.5983 Epoch 18/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9867 - accuracy: 0.6005 - val_loss: 0.9838 - val_accuracy: 0.5896 Epoch 19/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9867 - accuracy: 0.6000 - val_loss: 0.9795 - val_accuracy: 0.5955 Epoch 20/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9864 - accuracy: 0.5998 - val_loss: 0.9812 - val_accuracy: 0.5994 Epoch 21/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9864 - accuracy: 0.5996 - val_loss: 0.9791 - val_accuracy: 0.5947 Epoch 22/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9863 - accuracy: 0.6000 - val_loss: 0.9855 - val_accuracy: 0.5924 Epoch 23/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9860 - accuracy: 0.6006 - val_loss: 0.9762 - val_accuracy: 0.5996 Epoch 24/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9861 - accuracy: 0.6000 - val_loss: 0.9758 - val_accuracy: 0.6005 Epoch 25/25 4385/4385 [==============================] - 13s 3ms/step - loss: 0.9858 - accuracy: 0.6003 - val_loss: 0.9776 - val_accuracy: 0.5968
预测
为验证集和训练集运行预测。
plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
[<matplotlib.lines.Line2D at 0x7f9fc0030b50>]
train_eval_result = model.evaluate(dict(train_df), train_df['Sentiment'])
validation_eval_result = model.evaluate(dict(validation_df), validation_df['Sentiment'])
print(f"Training set accuracy: {train_eval_result[1]}")
print(f"Validation set accuracy: {validation_eval_result[1]}")
4385/4385 [==============================] - 12s 3ms/step - loss: 0.9820 - accuracy: 0.6014 493/493 [==============================] - 1s 2ms/step - loss: 0.9776 - accuracy: 0.5968 Training set accuracy: 0.6013968586921692 Validation set accuracy: 0.5967608690261841
混淆矩阵
另一个非常有趣的统计数据（尤其对于多类问题而言）是混淆矩阵。混淆矩阵允许可视化显示正确和错误标记的样本的比例。我们可以很容易看出分类器出现了多大偏差，以及标签分布是否有意义。理想情况下，预测中的最大数值应沿对角线分布。
predictions = model.predict(dict(validation_df))
predictions = tf.argmax(predictions, axis=-1)
predictions
<tf.Tensor: shape=(15745,), dtype=int64, numpy=array([1, 1, 2, ..., 2, 2, 2])>
cm = tf.math.confusion_matrix(validation_df['Sentiment'], predictions)
cm = cm/cm.numpy().sum(axis=1)[:, tf.newaxis]
sns.heatmap(
cm, annot=True,
xticklabels=SENTIMENT_LABELS,
yticklabels=SENTIMENT_LABELS)
plt.xlabel("Predicted")
plt.ylabel("True")
Text(32.99999999999999, 0.5, 'True')
我们可以将以下代码粘贴到代码单元，然后执行该代码，从而轻松将预测值提交回 Kaggle：
test_predictions = model.predict(dict(test_df))
test_predictions = np.argmax(test_predictions, axis=-1)
result_df = test_df.copy()
result_df["Predictions"] = test_predictions
result_df.to_csv(
"predictions.csv",
columns=["Predictions"],
header=["Sentiment"])
kaggle.api.competition_submit("predictions.csv", "Submitted from Colab",
"sentiment-analysis-on-movie-reviews")
提交后，查看排行榜了解您的表现。