為什么 Tensorflow 不在 GPU 上運行，而 GPU 設備在 Z23EEEB4347BDD26BFCZ6B7EE9A3B755DD 中被識別？

Question

我在windows 10中安裝了TensorFlow 2.2.0和TensorFlow-gpu 2.2.0 。 另外，我安裝了CUDA Toolkit v10.1並將cuDNN 7.6.5文件復制到CUDA directories中。 我的 GPU 是NVIDIA GeForce 940 MX 。 另外，我在 windows 上設置了 CUDA 路徑。 當我通過以下代碼測試設備時， CPU和GPU都被識別：

from tensorflow.python.client import device_lib
device_lib.list_local_devices()

output 是：

[name: "/device:CPU:0"
 device_type: "CPU"
 memory_limit: 268435456
 locality {
 }
 incarnation: 13265748925766868529,
 name: "/device:XLA_CPU:0"
 device_type: "XLA_CPU"
 memory_limit: 17179869184
 locality {
 }
 incarnation: 14569071601529958377
 physical_device_desc: "device: XLA_CPU device",
 name: "/device:XLA_GPU:0"
 device_type: "XLA_GPU"
 memory_limit: 17179869184
 locality {
 }
 incarnation: 15045400394346252324
 physical_device_desc: "device: XLA_GPU device"]

但是，當我運行我的代碼時，這些代碼似乎只在 CPU 上運行。 此外，當我使用tf.test.is_gpu_available()測試 GPU 可用性時，無法識別 GPU 設備並顯示False value 。
或者當我們運行tf.config.list_physical_devices('GPU')時，會打印一個empty list或[] 。 當我運行tf.config.experimental.list_physical_devices()時，這三個physical devices顯示在一個列表中：

[PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'),
 PhysicalDevice(name='/physical_device:XLA_CPU:0', device_type='XLA_CPU'),
 PhysicalDevice(name='/physical_device:XLA_GPU:0', device_type='XLA_GPU')]

重要的是，當我運行tf.config.list_physical_devices('XLA_GPU')時，將打印： [PhysicalDevice(name='/physical_device:XLA_GPU:0', device_type='XLA_GPU')]

此外，當我們運行代碼時，任務管理器顯示CPU 使用了 96% 的能力，而 GPU 僅使用了 1% 的能力。

我們運行的代碼如下：

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense, Dropout, Bidirectional
from tensorflow.keras.callbacks import ModelCheckpoint, TensorBoard
from sklearn import preprocessing
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from yahoo_fin import stock_info as si
from collections import deque

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import time
import os
import random


# set seed, so we can get the same results after rerunning several times
np.random.seed(314)
tf.random.set_seed(314)
random.seed(314)


def load_data(ticker, n_steps=50, scale=True, shuffle=True, lookup_step=1, 
                test_size=0.2, feature_columns=['adjclose', 'volume', 'open', 'high', 'low']):
    # see if ticker is already a loaded stock from yahoo finance
    if isinstance(ticker, str):
        # load it from yahoo_fin library
        df = si.get_data(ticker)
    elif isinstance(ticker, pd.DataFrame):
        # already loaded, use it directly
        df = ticker
    # this will contain all the elements we want to return from this function
    result = {}
    # we will also return the original dataframe itself
    result['df'] = df.copy()
    # make sure that the passed feature_columns exist in the dataframe
    for col in feature_columns:
        assert col in df.columns, f"'{col}' does not exist in the dataframe."
    if scale:
        column_scaler = {}
        # scale the data (prices) from 0 to 1
        for column in feature_columns:
            scaler = preprocessing.MinMaxScaler()
            df[column] = scaler.fit_transform(np.expand_dims(df[column].values, axis=1))
            column_scaler[column] = scaler

        # add the MinMaxScaler instances to the result returned
        result["column_scaler"] = column_scaler
    # add the target column (label) by shifting by `lookup_step`
    df['future'] = df['adjclose'].shift(-lookup_step)
    # last `lookup_step` columns contains NaN in future column
    # get them before droping NaNs
    last_sequence = np.array(df[feature_columns].tail(lookup_step))
    # drop NaNs
    df.dropna(inplace=True)
    sequence_data = []
    sequences = deque(maxlen=n_steps)
    for entry, target in zip(df[feature_columns].values, df['future'].values):
        sequences.append(entry)
        if len(sequences) == n_steps:
            sequence_data.append([np.array(sequences), target])
    # get the last sequence by appending the last `n_step` sequence with `lookup_step` sequence
    # for instance, if n_steps=50 and lookup_step=10, last_sequence should be of 59 (that is 50+10-1) length
    # this last_sequence will be used to predict in future dates that are not available in the dataset
    last_sequence = list(sequences) + list(last_sequence)
    # shift the last sequence by -1
    last_sequence = np.array(pd.DataFrame(last_sequence).shift(-1).dropna())
    # add to result
    result['last_sequence'] = last_sequence
    # construct the X's and y's
    X, y = [], []
    for seq, target in sequence_data:
        X.append(seq)
        y.append(target)
    # convert to numpy arrays
    X = np.array(X)
    y = np.array(y)
    # reshape X to fit the neural network
    X = X.reshape((X.shape[0], X.shape[2], X.shape[1]))
    # split the dataset
    result["X_train"], result["X_test"], result["y_train"], result["y_test"] = train_test_split(X, y, test_size=test_size, shuffle=shuffle)
    # return the result
    return result


def create_model(sequence_length, units=256, cell=LSTM, n_layers=2, dropout=0.3,
                loss="mean_absolute_error", optimizer="rmsprop", bidirectional=False):
    model = Sequential()
    for i in range(n_layers):
        if i == 0:
            # first layer
            if bidirectional:
                model.add(Bidirectional(cell(units, return_sequences=True), input_shape=(None, sequence_length)))
            else:
                model.add(cell(units, return_sequences=True, input_shape=(None, sequence_length)))
        elif i == n_layers - 1:
            # last layer
            if bidirectional:
                model.add(Bidirectional(cell(units, return_sequences=False)))
            else:
                model.add(cell(units, return_sequences=False))
        else:
            # hidden layers
            if bidirectional:
                model.add(Bidirectional(cell(units, return_sequences=True)))
            else:
                model.add(cell(units, return_sequences=True))
        # add dropout after each layer
        model.add(Dropout(dropout))
    model.add(Dense(1, activation="linear"))
    model.compile(loss=loss, metrics=["mean_absolute_error"], optimizer=optimizer)
    return model

# Window size or the sequence length
N_STEPS = 100
# Lookup step, 1 is the next day
LOOKUP_STEP = 1
# test ratio size, 0.2 is 20%
TEST_SIZE = 0.2
# features to use
FEATURE_COLUMNS = ["adjclose", "volume", "open", "high", "low"]
# date now
date_now = time.strftime("%Y-%m-%d")
### model parameters
N_LAYERS = 3
# LSTM cell
CELL = LSTM
# 256 LSTM neurons
UNITS = 256
# 40% dropout
DROPOUT = 0.4
# whether to use bidirectional RNNs
BIDIRECTIONAL = False
### training parameters
# mean absolute error loss
# LOSS = "mae"
# huber loss
LOSS = "huber_loss"
OPTIMIZER = "adam"
BATCH_SIZE = 64
EPOCHS = 400
# Apple stock market
ticker = "AAPL"
ticker_data_filename = os.path.join("data", f"{ticker}_{date_now}.csv")
# model name to save, making it as unique as possible based on parameters
model_name = f"{date_now}_{ticker}-{LOSS}-{OPTIMIZER}-{CELL.__name__}-seq-{N_STEPS}-step-{LOOKUP_STEP}-layers-{N_LAYERS}-units-{UNITS}"
if BIDIRECTIONAL:
    model_name += "-b"
    
# create these folders if they does not exist
if not os.path.isdir("results"):
    os.mkdir("results")
if not os.path.isdir("logs"):
    os.mkdir("logs")
if not os.path.isdir("data"):
    os.mkdir("data")


# load the data
data = load_data(ticker, N_STEPS, lookup_step=LOOKUP_STEP, test_size=TEST_SIZE, feature_columns=FEATURE_COLUMNS)

# save the dataframe
data["df"].to_csv(ticker_data_filename)

# construct the model
model = create_model(N_STEPS, loss=LOSS, units=UNITS, cell=CELL, n_layers=N_LAYERS,
                    dropout=DROPOUT, optimizer=OPTIMIZER, bidirectional=BIDIRECTIONAL)

# some tensorflow callbacks
checkpointer = ModelCheckpoint(os.path.join("results", model_name + ".h5"), save_weights_only=True, save_best_only=True, verbose=1)
tensorboard = TensorBoard(log_dir=os.path.join("logs", model_name))

history = model.fit(data["X_train"], data["y_train"],
                    batch_size=BATCH_SIZE,
                    epochs=EPOCHS,
                    validation_data=(data["X_test"], data["y_test"]),
                    callbacks=[checkpointer, tensorboard],
                    verbose=1)

model.save(os.path.join("results", model_name) + ".h5")

**GPU and CPU performance**如下：

請問你能幫幫我嗎？

Answer 1

我用 set NVIDIA Control Panel解決了這個問題。 我在桌面上按右鍵並選擇NVIDIA Control panel ：

---

然后，通過Set PhysX Configuration ，我 go 到Select a PhysX Processor和 select Auto-Select recommended如下：

---

此外，從Manage 3D settings ，我通過單擊Restore button恢復了設置： 此外，您可以從本節的Program Settings中設置 GPU 上的 Python。 我做到了。請在每個階段應用所有更改。 最后，運行上述每個代碼的有利結果如下所示：

---

Code1:
from tensorflow.python.client import device_lib
device_lib.list_local_devices()
Output1:
[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
incarnation: 12330560057435677891
, name: "/device:XLA_CPU:0"
device_type: "XLA_CPU"
memory_limit: 17179869184
locality {
}
incarnation: 14076398930644318194
physical_device_desc: "device: XLA_CPU device"
, name: "/device:GPU:0"
device_type: "GPU"
memory_limit: 3186897715
locality {
  bus_id: 1
  links {
  }
}
incarnation: 5889399188264267952
physical_device_desc: "device: 0, name: GeForce 940MX, pci bus id: 0000:01:00.0, compute capability: 5.0"
, name: "/device:XLA_GPU:0"
device_type: "XLA_GPU"
memory_limit: 17179869184
locality {
}
incarnation: 8080361800351872259
physical_device_desc: "device: XLA_GPU device"
]

---

Code2:
import tensorflow as tf
tf.config.list_physical_devices('GPU')

Output2:
[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

---

Code3:
tf.test.is_gpu_available()
Output3:
True

---

Answer 2

代碼應該是這樣的：

def load_data(ticker, n_steps=N_STEPS)

所以你可以在代碼中傳遞你定義為 100 的變量。