使用Android预先构建的tensorflow tflight模型(来自Python代码中的Android示例)

bounty将在3天后过期。回答此问题可获得+50的声望奖励。Hossam Hassan正在寻找来自知名来源的答案。

我最近下载了一个由TensorFlow制作的用于文本分类的Android演示应用，这是GitHub链接android GitHub demo
在该应用程序中，它有两种模式来预测文本是正面还是负面，使用（AverageWordVec / MobileBERT），但MobileBERT准确性更高
现在，我试图搜索替代，但使用Python，但我找不到任何使用相同模式，甚至接近相同的accurecy。
有没有办法在python中使用mobilebert.tflite文件并得到同样的预测？如果有人能给予我在python中运行该文件的示例代码，我将非常感激

更新

当我使用tokenizer方法时，它运行，但我得到以下输出：

2022-11-15 15:22:40.048025: I tensorflow/core/platform/cpu_feature_guard.cc:193] 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.
2022-11-15 15:22:40.177014: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory
2022-11-15 15:22:40.177060: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2022-11-15 15:22:40.208395: E tensorflow/stream_executor/cuda/cuda_blas.cc:2981] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2022-11-15 15:22:40.831937: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory
2022-11-15 15:22:40.832028: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory
2022-11-15 15:22:40.832041: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.
2022-11-15 15:22:41.897720: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcuda.so.1'; dlerror: libcuda.so.1: cannot open shared object file: No such file or directory
2022-11-15 15:22:41.897765: W tensorflow/stream_executor/cuda/cuda_driver.cc:263] failed call to cuInit: UNKNOWN ERROR (303)
2022-11-15 15:22:41.897783: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:156] kernel driver does not appear to be running on this host (Android-CI-CD): /proc/driver/nvidia/version does not exist
2022-11-15 15:22:41.898114: I tensorflow/core/platform/cpu_feature_guard.cc:193] 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.
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.

并且当我使用tflite_support.task方法时，我会得到以下输出

2022-11-15 15:25:30.210277: I tensorflow/core/platform/cpu_feature_guard.cc:193] 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.
2022-11-15 15:25:30.342254: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory
2022-11-15 15:25:30.342295: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2022-11-15 15:25:30.372550: E tensorflow/stream_executor/cuda/cuda_blas.cc:2981] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2022-11-15 15:25:30.990440: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory
2022-11-15 15:25:30.990522: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory
2022-11-15 15:25:30.990532: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Segmentation fault (core dumped)

**方法1：**您可以透过安装pip install tflite-support来使用tflite-support。

from tflite_support.task import text
classifier = text.BertNLClassifier.create_from_file('mobilebert.tflite')
classifier.classify('go to hell')

#Output:
[Classifications(categories=[Category(index=0, score=0.9992809891700745, display_name='', category_name='negative'), Category(index=0, score=0.0007190419128164649, display_name='', category_name='positive')], head_index=0, head_name='')])

方法2：

获取标记化器并设置推理：

from transformers import MobileBertTokenizer #pip install transformers

tokenizer = MobileBertTokenizer.from_pretrained("google/mobilebert-uncased")

inputs = tokenizer('go to hell!', return_tensors="tf")

# Load TFLite model and allocate tensors.
interpreter = tf.lite.Interpreter(model_path="mobilebert.tflite")

# Get input and output tensors.
input_details = interpreter.get_input_details()
output_details = interpreter.get_output_details()

interpreter.allocate_tensors()

运行推断：

#set input
input_shape = input_details[0]['shape']
word_ids = tf.keras.preprocessing.sequence.pad_sequences(inputs['input_ids'], maxlen=128, padding='post')
input_type_ids =  tf.keras.preprocessing.sequence.pad_sequences(inputs['token_type_ids'], maxlen=128, padding='post')
mask = tf.keras.preprocessing.sequence.pad_sequences(inputs['attention_mask'], maxlen=128, padding='post')

interpreter.set_tensor(input_details[0]['index'], word_ids)
interpreter.set_tensor(input_details[1]['index'], input_type_ids)
interpreter.set_tensor(input_details[2]['index'], mask)

# run the inference
interpreter.invoke()

#output
interpreter.get_tensor(output_details[0]['index'])

使用Android预先构建的tensorflow tflight模型(来自Python代码中的Android示例)

1条答案

相关问题

热门标签

最新问答