Deep learning model predicting same output for every input - deep-learning

My model should be able to multi-label each {image, text} into 4 classes with each class having {4,4,4,2} subclasses respectively. So for each input, the output should be a subclass of each class.
Output should be of length 4.
I'm labelling the images, with the input as {image, text content in the image} So the model should take input as both text-content present in the image, image data.
To process the image data, I'm importing resnet50 as my base model.
base_model_1 = tf.keras.applications.ResNet50(input_shape=X[0].shape,
include_top=False,
weights='imagenet')
base_model_1.trainable = True
Later joined few other layers to the base_model_1, so I can merge the image_model with text_model.
Image model be like :
def image_model():
image_input = tf.keras.Input(shape=(100, 100, 3), name = 'image_input')
image_layers = preprocess_input(image_input)
layer_bm_1 = base_model_1(image_layers, training=True)
dropout_layer = Dropout(0.2)(layer_bm_1)
layer_bm_1 = Conv2D(2048, kernel_size=3,padding='valid')(image_layers)
dropout_layer = Dropout(0.3)(layer_bm_1)
layer_bm_1 = Dense(1024)(layer_bm_1)
dropout_layer = Dropout(0.5)(layer_bm_1)
image_layers = GlobalAveragePooling2D()(layer_bm_1)
image_layers = Dropout(0.5, name = 'dropout_layer')(image_layers)
return image_input, image_layers
Text model be like :
def text_model():
text_input = tf.keras.Input(shape=(1,),dtype=tf.string, name='text')
text_layers = vectorize_layer(text_input)
text_layers = tf.keras.layers.Embedding(vocab_size, embedding_dim,input_length= sequence_length , name="embedding")(text_layers)
text_layers = tf.keras.layers.LSTM(1024, activation='relu', return_sequences=False)(text_layers)
text_layers = tf.keras.layers.Dense(1024, activation="relu")(text_layers)
#text_layers = tf.keras.layers.GlobalAveragePooling1D()(text_layers)
return text_input, text_layers
The output of both layers are concatenated.
def model(image_layer, text_layer, image_input, text_input):
concatenate = tf.keras.layers.concatenate([layer_1, layer_2])
semi_final_layer = tf.keras.layers.Dense(2048, activation='relu')(concatenate)
semi_final_layer = tf.keras.layers.Dense(1024, activation='relu')(semi_final_layer)
prediction_layer_1 = tf.keras.layers.Dense(516, activation='relu')(semi_final_layer)
prediction_layer_2 = tf.keras.layers.Dense(516, activation='relu')(semi_final_layer)
prediction_layer_3 = tf.keras.layers.Dense(516, activation='relu')(semi_final_layer)
prediction_layer_4 = tf.keras.layers.Dense(516, activation='relu')(semi_final_layer)
prediction_layer_1 = tf.keras.layers.Dense(254, activation='relu')(prediction_layer_1)
prediction_layer_2 = tf.keras.layers.Dense(254, activation='relu')(prediction_layer_2)
prediction_layer_3 = tf.keras.layers.Dense(254, activation='relu')(prediction_layer_3)
prediction_layer_4 = tf.keras.layers.Dense(254, activation='relu')(prediction_layer_4)
output_1 = tf.keras.layers.Dense(4, activation='softmax', name = 'humour')(prediction_layer_1)
output_2 = tf.keras.layers.Dense(4, activation='softmax', name = 'sarcasm')(prediction_layer_2)
output_3 = tf.keras.layers.Dense(4, activation='softmax', name = 'offensive')(prediction_layer_3)
output_4 = tf.keras.layers.Dense(2, activation='softmax', name = 'motivational')(prediction_layer_4)
model = tf.keras.Model(inputs = [image_input, text_input] ,
outputs = [output_1, output_2, output_3, output_4])
return model
# Function for decaying the learning rate.
def decay(epoch):
if epoch < 3:
return 1e-4
elif epoch >= 3 and epoch < 5:
return 1e-5
else:
return 1e-6
# Callback for printing the LR at the end of each epoch.
class PrintLR(tf.keras.callbacks.Callback):
def on_epoch_end(self, epoch, logs=None):
print('\nLearning rate for epoch {} is {}'.format(epoch + 1,
model.optimizer.lr.numpy()))
callbacks = [
tf.keras.callbacks.TensorBoard(log_dir='./logs'),
tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_prefix,
save_weights_only=True),
tf.keras.callbacks.LearningRateScheduler(decay),
PrintLR()
]
After training the above model with batchsize = 32, epochs = 10.
Total dataset contain = 7000 datapoints, splitted into training, validation, testing data in 0.7, 0.2, 0.1 ratio respectively.
Training, Validation accuracy, loss staying the same:
Epoch 1:
loss: 4.3440 - humour_loss: 1.3068 - sarcasm_loss: 1.1911 - offensive_loss: 1.1937 - motivational_loss: 0.6525 - humour_categorical_accuracy: 0.3384 - sarcasm_categorical_accuracy: 0.5011 - offensive_categorical_accuracy: 0.3709 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2886 - val_humour_loss: 1.2852 - val_sarcasm_loss: 1.1741 - val_offensive_loss: 1.1755 - val_motivational_loss: 0.6537 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-04
Epoch 2/10
loss: 4.3215 - humour_loss: 1.3028 - sarcasm_loss: 1.1843 - offensive_loss: 1.1839 - motivational_loss: 0.6504 - humour_categorical_accuracy: 0.3491 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3769 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2617 - val_humour_loss: 1.2763 - val_sarcasm_loss: 1.1710 - val_offensive_loss: 1.1675 - val_motivational_loss: 0.6469 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3832 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-04
Epoch 3/10
loss: 4.3259 - humour_loss: 1.3029 - sarcasm_loss: 1.1861 - offensive_loss: 1.1851 - motivational_loss: 0.6519 - humour_categorical_accuracy: 0.3409 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3744 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2731 - val_humour_loss: 1.2760 - val_sarcasm_loss: 1.1784 - val_offensive_loss: 1.1723 - val_motivational_loss: 0.6465 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-04
Epoch 4/10
loss: 4.3079 - humour_loss: 1.2993 - sarcasm_loss: 1.1796 - offensive_loss: 1.1799 - motivational_loss: 0.6491 - humour_categorical_accuracy: 0.3458 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3756 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2787 - val_humour_loss: 1.2806 - val_sarcasm_loss: 1.1780 - val_offensive_loss: 1.1729 - val_motivational_loss: 0.6472 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-05
Epoch 5/10
loss: 4.3078 - humour_loss: 1.2990 - sarcasm_loss: 1.1808 - offensive_loss: 1.1796 - motivational_loss: 0.6485 - humour_categorical_accuracy: 0.3456 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3782 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2705 - val_humour_loss: 1.2787 - val_sarcasm_loss: 1.1744 - val_offensive_loss: 1.1708 - val_motivational_loss: 0.6467 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-05
Epoch 6/10
loss: 4.3078 - humour_loss: 1.2995 - sarcasm_loss: 1.1803 - offensive_loss: 1.1798 - motivational_loss: 0.6482 - humour_categorical_accuracy: 0.3456 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3733 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2710 - val_humour_loss: 1.2790 - val_sarcasm_loss: 1.1745 - val_offensive_loss: 1.1709 - val_motivational_loss: 0.6467 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-06
Epoch 7/10
loss: 4.3059 - humour_loss: 1.2983 - sarcasm_loss: 1.1792 - offensive_loss: 1.1798 - motivational_loss: 0.6486 - humour_categorical_accuracy: 0.3453 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3805 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2701 - val_humour_loss: 1.2786 - val_sarcasm_loss: 1.1742 - val_offensive_loss: 1.1706 - val_motivational_loss: 0.6467 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-06
Epoch 8/10
loss: 4.3060 - humour_loss: 1.2986 - sarcasm_loss: 1.1793 - offensive_loss: 1.1796 - motivational_loss: 0.6485 - humour_categorical_accuracy: 0.3456 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3765 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2701 - val_humour_loss: 1.2786 - val_sarcasm_loss: 1.1742 - val_offensive_loss: 1.1706 - val_motivational_loss: 0.6467 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-06
Epoch 9/10
loss: 4.3048 - humour_loss: 1.2987 - sarcasm_loss: 1.1789 - offensive_loss: 1.1792 - motivational_loss: 0.6480 - humour_categorical_accuracy: 0.3456 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3767 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2698 - val_humour_loss: 1.2785 - val_sarcasm_loss: 1.1742 - val_offensive_loss: 1.1705 - val_motivational_loss: 0.6466 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-06
Epoch 10/10
loss: 4.3065 - humour_loss: 1.2980 - sarcasm_loss: 1.1804 - offensive_loss: 1.1796 - motivational_loss: 0.6484 - humour_categorical_accuracy: 0.3456 - sarcasm_categorical_accuracy: 0.5034 - offensive_categorical_accuracy: 0.3827 - motivational_categorical_accuracy: 0.6486 - val_loss: 4.2702 - val_humour_loss: 1.2786 - val_sarcasm_loss: 1.1743 - val_offensive_loss: 1.1706 - val_motivational_loss: 0.6467 - val_humour_categorical_accuracy: 0.3885 - val_sarcasm_categorical_accuracy: 0.4978 - val_offensive_categorical_accuracy: 0.3975 - val_motivational_categorical_accuracy: 0.6517 - lr: 1.0000e-06
The output for a set of inputs from test_Data is :
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
[0.0219533 0.06215543 0.3258566 0.46769002 0.12721306]
The above output is the same for every test_input.
Let me know any ideas to tackle the issue here.
Edit 1 :
input shape for image : (100,100,3)

Related

Why does the reward of my model not pass a threshold?

The problem is that at the beginning the reward increases its value up to -400, as it trains it decreases up to -100, but from there it does not continue increasing towards zero, it stays there stagnant.
10000/10000 [==============================] - 111s 11ms/step - reward: -100.0000
Interval 2 (10000 steps performed)
10000/10000 [==============================] - 126s 13ms/step - reward: -100.0000
Interval 3 (20000 steps performed)
10000/10000 [==============================] - 120s 12ms/step - reward: -100.0000
Interval 4 (30000 steps performed)
10000/10000 [==============================] - 126s 13ms/step - reward: -100.0000
Interval 5 (40000 steps performed)
9049/10000 [==========================>...] - ETA: 11s - reward: -100.0000
The model:
def step(self, action):
reward = 0
paqueteSale = False
listaPuertos, cargaPuertos, nodoSrc, nodoDst, tamPaquete = self.traducirEstado()
#Here are the different actions
#Calculamos la recompensa
if paqueteSale:
if calcularMLUGlobal() < umbralMLU:
reward = 1
else:
reward = umbralMLU-calcularMLUGlobal()
else:
reward = -100
self.rellenarEstado(listaPuertos, cargaPuertos)
#Devovlemos la informacion
return self.estado, reward, False, {}
Does anyone have any idea why this is happening? I am using DQNAgent from rl.agents for training.

How to tune the batch size for LSTM using hyperband?

I am trying to tune the batch size for LSTM using hyperband optimisation using the following codes, but it didn't work because the numbers of the training items did not change in different batch size conditions. Do you know how to improve the codes? Thank you in advance.
import keras_tuner as kt
import tensorflow as tf
from tensorflow import keras
import numpy as np
x_train = np.random.rand(63, 92)
y_train = np.random.randint(0,6, (63))
x_val = np.random.rand(63, 92)
y_val = np.random.randint(0,7, (63))
train_set = tf.keras.preprocessing.timeseries_dataset_from_array(
x_train, y_train, sequence_length=10)
val_set = tf.keras.preprocessing.timeseries_dataset_from_array(
x_val, y_val, sequence_length=10)
train_set = tf.keras.preprocessing.timeseries_dataset_from_array(
x_train, y_train, sequence_length=10, batch_size =1)
val_set = tf.keras.preprocessing.timeseries_dataset_from_array(
x_val, y_val, sequence_length=10, batch_size =1)
def model_builder(hp):
lr = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
hp_units = hp.Int('units', min_value=32, max_value=256, step=32)
hp_units1 = hp.Int('units1', min_value=32, max_value=256, step=32)
lstm_model = tf.keras.models.Sequential([
tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(hp_units, return_sequences=True)),
tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(hp_units1)),
tf.keras.layers.Dense(units=1)
])
lstm_model.compile(loss='mse',
optimizer=tf.optimizers.Adam(learning_rate=lr),
metrics=['mse'])
return lstm_model
class MyTuner(kt.tuners.Hyperband):
def run_trial(self, trial, *args, **kwargs):
kwargs['batch_size'] = trial.hyperparameters.Int('batch_size', 2, 6, step=2)
return super(MyTuner, self).run_trial(trial, *args, **kwargs)
tuner = MyTuner(model_builder,
objective='val_loss',
max_epochs=4,
factor=3,
directory='KT',
project_name='intro_to_kt0207',
overwrite=True)
early_stop = tf.keras.callbacks.EarlyStopping(
monitor='val_loss', min_delta=0, patience=20, verbose=0,
mode='min', baseline=None, restore_best_weights=True
)
tuner.search(train_set, epochs=1000,
validation_data = val_set)
This is the output:
Search: Running Trial #2
Hyperparameter |Value |Best Value So Far
learning_rate |0.0001 |0.01
units |192 |224
units1 |192 |64
batch_size |4 |2
tuner/epochs |2 |2
tuner/initial_e...|0 |0
tuner/bracket |1 |1
tuner/round |0 |0
Epoch 1/2
54/54 [==============================] - 7s 37ms/step - loss: 4.4262 - mse: 4.4262 - val_loss: 3.4221 - val_mse: 3.4221
Epoch 2/2
54/54 [==============================] - 1s 12ms/step - loss: 3.1213 - mse: 3.1213 - val_loss: 3.4463 - val_mse: 3.4463
Trial 2 Complete [00h 00m 08s]
val_loss: 3.42207670211792
Best val_loss So Far: 3.1889588832855225
Total elapsed time: 00h 00m 18s
Search: Running Trial #3
Hyperparameter |Value |Best Value So Far
learning_rate |0.01 |0.01
units |192 |224
units1 |96 |64
batch_size |2 |2
tuner/epochs |2 |2
tuner/initial_e...|0 |0
tuner/bracket |1 |1
tuner/round |0 |0
Epoch 1/2
54/54 [==============================] - 7s 34ms/step - loss: 3.6241 - mse: 3.6241 - val_loss: 3.1699 - val_mse: 3.1699
Epoch 2/2
54/54 [==============================] - 1s 12ms/step - loss: 3.1807 - mse: 3.1807 - val_loss: 3.2480 - val_mse: 3.2480
Trial 3 Complete [00h 00m 08s]
val_loss: 3.1699421405792236
Best val_loss So Far: 3.1699421405792236
Total elapsed time: 00h 00m 26s

training vgg on flowers dataset with keras, validation loss not changing

I am doing a little experiment on VGG network with keras.
The dataset I use is the flowers dataset with 5 classes including rose, sunflower, dandelion, tulip and daisy.
There is something I could not figure out:
When I used a small CNN network(not VGG, in the code below), it converged quickly and reached a validation accuracy about 75% after only about 8 epochs.
Then I switched to VGG network(the commented out area in the code). The loss and accuracy of the network just did not change at all, it output something like:
Epoch 1/50 402/401 [==============================] - 199s 495ms/step - loss: 13.3214 - acc: 0.1713 - val_loss: 13.0144 - val_acc: 0.1926
Epoch 2/50
402/401 [==============================] - 190s 473ms/step - loss: 13.3473 - acc: 0.1719 - val_loss: 13.0144 - val_acc: 0.1926
Epoch 3/50
402/401 [==============================] - 204s 508ms/step - loss: 13.3423 - acc: 0.1722 - val_loss: 13.0144 - val_acc: 0.1926
Epoch 4/50
402/401 [==============================] - 190s 472ms/step - loss: 13.3522 - acc: 0.1716 - val_loss: 13.0144 - val_acc: 0.1926
Epoch 5/50
402/401 [==============================] - 189s 471ms/step - loss: 13.3364 - acc: 0.1726 - val_loss: 13.0144 - val_acc: 0.1926
Epoch 6/50
402/401 [==============================] - 189s 471ms/step - loss: 13.3453 - acc: 0.1720 - val_loss: 13.0144 - val_acc: 0.1926
Epoch 7/50
Epoch 7/50 402/401 [==============================] - 189s 471ms/step - loss: 13.3503 - acc: 0.1717 - val_loss: 13.0144 - val_acc: 0.1926
PS: I did this experiment with other datasets and frameworks as well (place365 dataset with tensorflow and slim). The result is just the same. I have looked into the VGG paper(Simonyan&Zisserman), it says there are multiple stages to train a deep network like VGG, like from stage A to stage E with different network structures. I am not sure I have to train my VGG network the same way as it is described in the VGG paper. And other online courses did not mention this complex training process as well.
Anyone has any ideas?
My code:
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Activation, Dropout, Flatten, Dense
from keras import backend as K
# dimensions of our images.
img_width, img_height = 224, 224
train_data_dir = './data/train'
validation_data_dir = './data/val'
nb_train_samples = 3213
nb_validation_samples = 457
epochs = 50
batch_size = 8
if K.image_data_format() == 'channels_first':
input_shape = (3, img_width, img_height)
else:
input_shape = (img_width, img_height, 3)
# random cnn model:
model = Sequential()
model.add(Conv2D(32, (3, 3), input_shape=input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Flatten())
model.add(Dense(64))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(5))
model.add(Activation('softmax'))
# vgg model:
'''model = Sequential([
Conv2D(64, (3, 3), input_shape=input_shape, padding='same',
activation='relu'),
Conv2D(64, (3, 3), activation='relu', padding='same'),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(128, (3, 3), activation='relu', padding='same'),
Conv2D(128, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(256, (3, 3), activation='relu', padding='same',),
Conv2D(256, (3, 3), activation='relu', padding='same',),
Conv2D(256, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
Conv2D(512, (3, 3), activation='relu', padding='same',),
MaxPooling2D(pool_size=(2, 2), strides=(2, 2)),
Flatten(),
Dense(256, activation='relu'),
Dense(256, activation='relu'),
Dense(5, activation='softmax')
])'''
model.compile(loss='categorical_crossentropy',
optimizer='rmsprop',
metrics=['accuracy'])
train_datagen = ImageDataGenerator(
rescale=1. / 255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1. / 255)
train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size=(img_width, img_height),
batch_size=batch_size,
class_mode='categorical')
validation_generator = test_datagen.flow_from_directory(
validation_data_dir,
target_size=(img_width, img_height),
batch_size=batch_size,
class_mode='categorical')
model.fit_generator(
train_generator,
steps_per_epoch=nb_train_samples // batch_size,
epochs=epochs,
validation_data=validation_generator,
validation_steps=nb_validation_samples // batch_size)
model.save_weights('flowers.h5')

Problem solved, I changed my learning rate to 0.0001.
It starts to learn now.
It seems like 0.001 is not small enough.

Select id where all user id match

Okay, I get an array with multiple user ids from a php script.
Now I have this mysql table:
+ - - - - - - - - - - - + - - - - - +
¦ fk_conversation ¦ fk_user ¦
+ - - - - - - - - - - - + - - - - - +
¦ 1 ¦ 2 ¦
+ - - - - - - - - - - - + - - - - - +
¦ 1 ¦ 3 ¦
+ - - - - - - - - - - - + - - - - - +
¦ 1 ¦ 4 ¦
+ - - - - - - - - - - - + - - - - - +
¦ 2 ¦ 2 ¦
+ - - - - - - - - - - - + - - - - - +
¦ 2 ¦ 4 ¦
+ - - - - - - - - - - - + - - - - - +
If I get the array (2,4) I need to get the fk_conversation key = 2 but if I get the array(2,3,4) I should get the fk_conversation key = 1.
I always want to get the fk_conversation id where all connected fk_user ids are in the array.

select fk_conversation
from my_table
where fk_user in ( ARRAY VALUES )
and fk_conversation in (
select fk_conversation
from my_table
group by fk_conversation having count(distinct fk_user) = ARRAY SIZE
)
group by fk_conversation
having count(distinct fk_user) = ARRAY SIZE;

Try:
select fk_conversation
from table
where fk_user in ( comma separated array_values )
group by fk_conversation
having count(fk_conversation) = array_size

App crashes after launching on OS 3.1

I need help to understand the crash log i received from the client. The app works fine on my phone but it crashes with him. I have OS 3.0 installed here while my client has upgraded to OS 3.1.
Client reported app crashes usually when he starts the app.
Why I am getting EXC_CRASH (SIGABRT)? Can anybody point me in the right direction here?
Incident Identifier: AA3A4B6A-BD0D-473F-B1D2-EF9655A0116E
CrashReporter Key: f3a4736dc8d450a3cb0ecb7367313dbbd816c484
Process: MyApp [730]
Path: /var/mobile/Applications/D0F3BF28-B790-4CE4-8A40-08577C4B34C9/MyApp_36.app/MyApp
Identifier: MyApp
Version: ??? (???)
Code Type: ARM (Native)
Parent Process: launchd [1]
Date/Time: 2009-10-05 17:10:07.101 -0600
OS Version: iPhone OS 3.1 (7C144)
Report Version: 104
Exception Type: EXC_CRASH (SIGABRT)
Exception Codes: 0x00000000, 0x00000000
Crashed Thread: 0
Thread 0 Crashed:
0 libSystem.B.dylib 0x3080c4b8 0x3080b000 + 5304
1 libSystem.B.dylib 0x3080f094 0x3080b000 + 16532
2 CoreFoundation 0x3147a002 0x31423000 + 356354
3 CoreFoundation 0x31479c18 0x31423000 + 355352
4 GraphicsServices 0x325323a0 0x3252e000 + 17312
5 UIKit 0x3314ec28 0x3314b000 + 15400
6 UIKit 0x3314d228 0x3314b000 + 8744
7 MyApp 0x00002090 main (main.m:13)
8 MyApp 0x0000202c start + 44
Thread 1:
0 libSystem.B.dylib 0x3080c4b8 0x3080b000 + 5304
1 libSystem.B.dylib 0x3080f094 0x3080b000 + 16532
2 CoreFoundation 0x3147a002 0x31423000 + 356354
3 CoreFoundation 0x31479c18 0x31423000 + 355352
4 WebCore 0x31e196f0 0x31d95000 + 542448
5 libSystem.B.dylib 0x308367b0 0x3080b000 + 178096
Thread 2:
0 libSystem.B.dylib 0x3080c4b8 0x3080b000 + 5304
1 libSystem.B.dylib 0x3080f094 0x3080b000 + 16532
2 CoreFoundation 0x3147a002 0x31423000 + 356354
3 CoreFoundation 0x31479c18 0x31423000 + 355352
4 Foundation 0x31382998 0x31328000 + 371096
5 Foundation 0x3137bac6 0x31328000 + 342726
6 Foundation 0x31329d0e 0x31328000 + 7438
7 libSystem.B.dylib 0x308367b0 0x3080b000 + 178096
Thread 3:
0 libSystem.B.dylib 0x308312f0 0x3080b000 + 156400
1 CoreFoundation 0x314437e2 0x31423000 + 133090
2 libSystem.B.dylib 0x308367b0 0x3080b000 + 178096
Thread 4:
0 libSystem.B.dylib 0x3089c720 0x3080b000 + 595744
1 libSystem.B.dylib 0x3083e4d8 0x3080b000 + 210136
2 libSystem.B.dylib 0x3083dc9c 0x3080b000 + 208028
3 PhotoLibrary 0x32f545b8 0x32f11000 + 275896
4 Foundation 0x3137bac6 0x31328000 + 342726
5 Foundation 0x31329d0e 0x31328000 + 7438
6 libSystem.B.dylib 0x308367b0 0x3080b000 + 178096
Thread 5:
0 libSystem.B.dylib 0x3089c720 0x3080b000 + 595744
1 libSystem.B.dylib 0x3083e4d8 0x3080b000 + 210136
2 libSystem.B.dylib 0x3083dc9c 0x3080b000 + 208028
3 CoreMedia 0x30635660 0x30630000 + 22112
4 CoreMedia 0x3063553c 0x30630000 + 21820
5 MediaToolbox 0x30bc0bb4 0x30bbd000 + 15284
6 libSystem.B.dylib 0x308367b0 0x3080b000 + 178096
Thread 6:
0 libSystem.B.dylib 0x3080db48 0x3080b000 + 11080
1 libSystem.B.dylib 0x3080d9b0 0x3080b000 + 10672
2 CoreFoundation 0x31427bc4 0x31423000 + 19396
3 CoreFoundation 0x3145f2ee 0x31423000 + 246510
4 CoreFoundation 0x3145bb90 0x31423000 + 232336
5 CoreFoundation 0x3145b9c0 0x31423000 + 231872
6 Foundation 0x3136d29c 0x31328000 + 283292
7 Foundation 0x3136f990 0x31328000 + 293264
8 Foundation 0x31345668 0x31328000 + 120424
9 CoreFoundation 0x3149e8a2 0x31423000 + 506018
10 CoreFoundation 0x314612f4 0x31423000 + 254708
11 CoreFoundation 0x3145feca 0x31423000 + 249546
12 CoreFoundation 0x31496fd6 0x31423000 + 475094
13 CoreFoundation 0x3144c4e8 0x31423000 + 169192
14 CoreFoundation 0x314b7de2 0x31423000 + 609762
15 libobjc.A.dylib 0x3431288c 0x3430a000 + 34956
16 libstdc++.6.dylib 0x32d51a84 0x32cee000 + 408196
17 libstdc++.6.dylib 0x32d51afc 0x32cee000 + 408316
18 libstdc++.6.dylib 0x32d51c24 0x32cee000 + 408612
19 libobjc.A.dylib 0x34310e54 0x3430a000 + 28244
20 CoreFoundation 0x31449b2c 0x31423000 + 158508
21 CoreFoundation 0x31449acc 0x31423000 + 158412
22 Foundation 0x3134970a 0x31328000 + 136970
23 Foundation 0x3136f520 0x31328000 + 292128
24 MyApp 0x0002c4aa -[BlogViewController tableView:heightForRowAtIndexPath:] (BlogViewController.m:239)
25 UIKit 0x331eae98 0x3314b000 + 655000
26 UIKit 0x331ea5a0 0x3314b000 + 652704
27 UIKit 0x33198428 0x3314b000 + 316456
28 UIKit 0x33160428 0x3314b000 + 87080
29 UIKit 0x33197e04 0x3314b000 + 314884
30 UIKit 0x331977b8 0x3314b000 + 313272
31 MyApp 0x0002c738 -[BlogViewController loadMessagesThread] (BlogViewController.m:135)
32 Foundation 0x3137bac6 0x31328000 + 342726
33 Foundation 0x31329d0e 0x31328000 + 7438
34 libSystem.B.dylib 0x308367b0 0x3080b000 + 178096
Thread 7:
0 libSystem.B.dylib 0x3089bb5c 0x3080b000 + 592732
1 libSystem.B.dylib 0x3089bb4a 0x3080b000 + 592714
2 libSystem.B.dylib 0x3089bb3e 0x3080b000 + 592702
3 libSystem.B.dylib 0x308b2e64 0x3080b000 + 687716
4 libstdc++.6.dylib 0x32d54390 0x32cee000 + 418704
5 libobjc.A.dylib 0x34312898 0x3430a000 + 34968
6 libstdc++.6.dylib 0x32d51a84 0x32cee000 + 408196
7 libstdc++.6.dylib 0x32d51afc 0x32cee000 + 408316
8 libstdc++.6.dylib 0x32d51c24 0x32cee000 + 408612
9 libobjc.A.dylib 0x34310e54 0x3430a000 + 28244
10 CoreFoundation 0x31449b2c 0x31423000 + 158508
11 CoreFoundation 0x31449acc 0x31423000 + 158412
12 Foundation 0x3134970a 0x31328000 + 136970
13 Foundation 0x3136f520 0x31328000 + 292128
14 MyApp 0x0002c4aa -[BlogViewController tableView:heightForRowAtIndexPath:] (BlogViewController.m:239)
15 UIKit 0x331eae98 0x3314b000 + 655000
16 UIKit 0x331ea5a0 0x3314b000 + 652704
17 UIKit 0x33198428 0x3314b000 + 316456
18 UIKit 0x33160428 0x3314b000 + 87080
19 UIKit 0x33197e04 0x3314b000 + 314884
20 UIKit 0x331977b8 0x3314b000 + 313272
21 MyApp 0x0002c738 -[BlogViewController loadMessagesThread] (BlogViewController.m:135)
22 Foundation 0x3137bac6 0x31328000 + 342726
23 Foundation 0x31329d0e 0x31328000 + 7438
24 libSystem.B.dylib 0x308367b0 0x3080b000 + 178096
Thread 0 crashed with ARM Thread State:
r0: 0x10004005 r1: 0x03000006 r2: 0x00000000 r3: 0x00000450
r4: 0x00001403 r5: 0x00000000 r6: 0x00000000 r7: 0x2ffff59c
r8: 0x00000000 r9: 0x038fc098 r10: 0x03000006 r11: 0x00000450
ip: 0xffffffe1 sp: 0x2ffff560 lr: 0x3080f09c pc: 0x3080c4b8
cpsr: 0x000f0010
Binary Images:
0x1000 - 0x3cfff +MyApp armv6 <7bfba5c63a8d3386a2f6773e22bed2e5> /var/mobile/Applications/D0F3BF28-B790-4CE4-8A40-08577C4B34C9/MyApp_36.app/MyApp
0xeb000 - 0xecfff dns.so armv6 <957f94410f77a351749ac39d2b4b4abe> /usr/lib/info/dns.so
0x2fe00000 - 0x2fe26fff dyld armv6 <c9ed2fd4b6ad0b603479d414cb382dd5> /usr/lib/dyld
0x30005000 - 0x30011fff CoreVideo armv6 <c31d33c7629e4f7b14a659f06f17674e> /System/Library/PrivateFrameworks/CoreVideo.framework/CoreVideo
0x30061000 - 0x30078fff OpenGLES armv6 <c31f4982bc65e67356e9b692f9976b1a> /System/Library/Frameworks/OpenGLES.framework/OpenGLES
0x30079000 - 0x301cdfff AudioToolbox armv6 <065ac06f6e9f6bc62555565581a50637> /System/Library/Frameworks/AudioToolbox.framework/AudioToolbox
0x30218000 - 0x30223fff libbz2.1.0.dylib armv6 <28bbad866f148b0d0adda83a01cec8f8> /usr/lib/libbz2.1.0.dylib
0x3035d000 - 0x3040afff ImageIO armv6 <8d44717aa7e4314848108d49b42ce576> /System/Library/PrivateFrameworks/ImageIO.framework/ImageIO
0x3040d000 - 0x30441fff Security armv6 <36f9d8ca8c4e037e14edd6ae95524b24> /System/Library/Frameworks/Security.framework/Security
0x3045f000 - 0x30493fff SystemConfiguration armv6 <818346d31133d475739fe3c3b35b5bd1> /System/Library/Frameworks/SystemConfiguration.framework/SystemConfiguration
0x30630000 - 0x30677fff CoreMedia armv6 <07c13fb5ab31a52dfc3a50be02898f1d> /System/Library/PrivateFrameworks/CoreMedia.framework/CoreMedia
0x3067b000 - 0x30687fff IAP armv6 <3752a21b994225b12069f929bf9b19d0> /System/Library/PrivateFrameworks/IAP.framework/IAP
0x30725000 - 0x3072dfff libgcc_s.1.dylib armv6 <6673f222acee5f3a208169712387862a> /usr/lib/libgcc_s.1.dylib
0x307a9000 - 0x307b7fff libz.1.dylib armv6 <c4d3d79ffaa7e13f556cd5f9c9fceff2> /usr/lib/libz.1.dylib
0x307b8000 - 0x30808fff Celestial armv6 <6ccb9601cfba52fc5b7740e70fb565b4> /System/Library/PrivateFrameworks/Celestial.framework/Celestial
0x30809000 - 0x3080afff CoreSurface armv6 <12f13b59faf0107d8536963b1552c481> /System/Library/PrivateFrameworks/CoreSurface.framework/CoreSurface
0x3080b000 - 0x30931fff libSystem.B.dylib armv6 <7bcb29dde047a859e6f57f2e9e63d38f> /usr/lib/libSystem.B.dylib
0x30986000 - 0x30988fff ArtworkCache armv6 <49b8fea9eb3f6e7d2241d94cc241bbd2> /System/Library/PrivateFrameworks/ArtworkCache.framework/ArtworkCache
0x309f6000 - 0x309fbfff ITSync armv6 <cad602d917d800f848e522d835c3a031> /System/Library/PrivateFrameworks/ITSync.framework/ITSync
0x30b45000 - 0x30b4bfff liblockdown.dylib armv6 <74ead45f4f6840457b982e3a5cc30055> /usr/lib/liblockdown.dylib
0x30b4d000 - 0x30b79fff CoreLocation armv6 <ac2b8ebbcb03f7a57c61c9a5eaa5b3d7> /System/Library/Frameworks/CoreLocation.framework/CoreLocation
0x30b7e000 - 0x30bbcfff VideoToolbox armv6 <15ea0e25a764b445ce59401f09c2787d> /System/Library/PrivateFrameworks/VideoToolbox.framework/VideoToolbox
0x30bbd000 - 0x30cd3fff MediaToolbox armv6 <4930d982e746b906ca6fac74e30d361b> /System/Library/PrivateFrameworks/MediaToolbox.framework/MediaToolbox
0x30eb7000 - 0x30eb9fff Camera armv6 <83d973c375886b5c4632beae0cea27c6> /System/Library/PrivateFrameworks/Camera.framework/Camera
0x30ef7000 - 0x30f16fff Bom armv6 <fcd9e839c0d1a25cbf59a08cc60fea3f> /System/Library/PrivateFrameworks/Bom.framework/Bom
0x30f1a000 - 0x30f1cfff CrashReporterSupport armv6 <db3c3a983db3d53f6199dafb8b472bfb> /System/Library/PrivateFrameworks/CrashReporterSupport.framework/CrashReporterSupport
0x31043000 - 0x3112afff MusicLibrary armv6 <3deebbf46222551b71d0121b6360308b> /System/Library/PrivateFrameworks/MusicLibrary.framework/MusicLibrary
0x311a4000 - 0x311abfff MobileMusicPlayer armv6 <5492f53cedd0e55e1fd504a087b65e08> /System/Library/PrivateFrameworks/MobileMusicPlayer.framework/MobileMusicPlayer
0x311ac000 - 0x31221fff MediaPlayer armv6 <8e2efee9a7b8abcfddec740e95c87ed3> /System/Library/Frameworks/MediaPlayer.framework/MediaPlayer
0x31222000 - 0x31224fff AppleJPEG armv6 <d28a5dc54781356536addeb4ed7235ab> /System/Library/PrivateFrameworks/AppleJPEG.framework/AppleJPEG
0x31225000 - 0x31322fff JavaScriptCore armv6 <303af93622bcb0c4b68ed62b55d124fe> /System/Library/PrivateFrameworks/JavaScriptCore.framework/JavaScriptCore
0x31328000 - 0x31420fff Foundation armv6 <8ddca22a76a5f2bc91f55c87eea8e9a5> /System/Library/Frameworks/Foundation.framework/Foundation
0x31423000 - 0x314cffff CoreFoundation armv6 <a40c65b893c6ca5c0b632d2804ad59d1> /System/Library/Frameworks/CoreFoundation.framework/CoreFoundation
0x314fc000 - 0x31556fff libsqlite3.dylib armv6 <e346522412f1e957d9cee6511c66bdc9> /usr/lib/libsqlite3.dylib
0x31557000 - 0x31557fff vecLib armv6 <c8a08b45effd5eb35a476a969628e52d> /System/Library/PrivateFrameworks/Accelerate.framework/Frameworks/vecLib.framework/vecLib
0x31558000 - 0x3155bfff MobileInstallation armv6 <e9617b1538fe22f2d6eb91156d130409> /System/Library/PrivateFrameworks/MobileInstallation.framework/MobileInstallation
0x3155c000 - 0x3156efff TelephonyUI armv6 <cf3ba0b4f4141796c0b7b376b48b4f20> /System/Library/PrivateFrameworks/TelephonyUI.framework/TelephonyUI
0x3158f000 - 0x315fcfff CFNetwork armv6 <19088cf3f0b4f667b52482ac6ab99690> /System/Library/Frameworks/CFNetwork.framework/CFNetwork
0x31603000 - 0x3160dfff libkxld.dylib armv6 <8c6d9927c48ad08cfae540d60af17b6a> /usr/lib/system/libkxld.dylib
0x3160e000 - 0x31617fff SpringBoardServices armv6 <a2849d569ba1e7a4083aec5be693f055> /System/Library/PrivateFrameworks/SpringBoardServices.framework/SpringBoardServices
0x31622000 - 0x3165ffff libCGFreetype.A.dylib armv6 <640dc156b068d347f6c83a27adad6d06> /System/Library/Frameworks/CoreGraphics.framework/Resources/libCGFreetype.A.dylib
0x31660000 - 0x3169efff CoreText armv6 <c21aa763bf62e82e8a1f062c3e201e84> /System/Library/PrivateFrameworks/CoreText.framework/CoreText
0x3169f000 - 0x316abfff DataAccessExpress armv6 <c16ab7cb09e768e26ea3fa714475b691> /System/Library/PrivateFrameworks/DataAccessExpress.framework/DataAccessExpress
0x316ac000 - 0x316bcfff libbsm.0.dylib armv6 <f1b3998fcf7b60cee74b3e03fc94bf31> /usr/lib/libbsm.0.dylib
0x317c5000 - 0x31872fff WebKit armv6 <b555f14fc6c288f8725843055d936507> /System/Library/PrivateFrameworks/WebKit.framework/WebKit
0x31d2b000 - 0x31d2bfff Accelerate armv6 <21e1dc9fad96f0d51afbeb7f7e006aaf> /System/Library/PrivateFrameworks/Accelerate.framework/Accelerate
0x31d95000 - 0x324e8fff WebCore armv6 <28b391254b0edba93084f458e5805787> /System/Library/PrivateFrameworks/WebCore.framework/WebCore
0x324f7000 - 0x3251efff AppSupport armv6 <d8d247ecb9f5a3ff92583048a5da264c> /System/Library/PrivateFrameworks/AppSupport.framework/AppSupport
0x3252e000 - 0x32539fff GraphicsServices armv6 <8629b6affa7c7752ab3c702f1c694325> /System/Library/PrivateFrameworks/GraphicsServices.framework/GraphicsServices
0x325f8000 - 0x326eefff libxml2.2.dylib armv6 <893355592f660060653383b29f42d102> /usr/lib/libxml2.2.dylib
0x326ef000 - 0x32729fff libvDSP.dylib armv6 <cf9cc6079374718dce29cba9de48b993> /System/Library/PrivateFrameworks/Accelerate.framework/Frameworks/vecLib.framework/libvDSP.dylib
0x32744000 - 0x3274cfff AccountSettings armv6 <9d0822c54adc0f81b866543fc34196bf> /System/Library/PrivateFrameworks/AccountSettings.framework/AccountSettings
0x32848000 - 0x32895fff IOKit armv6 <035ec9372089422af1ccdb17d72ac091> /System/Library/Frameworks/IOKit.framework/Versions/A/IOKit
0x328da000 - 0x328e9fff MobileBluetooth armv6 <6d87ea09ffe173f3d356514e7382bd95> /System/Library/PrivateFrameworks/MobileBluetooth.framework/MobileBluetooth
0x3290a000 - 0x329b4fff QuartzCore armv6 <cb65d19d96053b411b9a1c68f8f20a33> /System/Library/Frameworks/QuartzCore.framework/QuartzCore
0x329d6000 - 0x329f4fff libresolv.9.dylib armv6 <e2f1ba98b2a7b91e91896cea1db932e4> /usr/lib/libresolv.9.dylib
0x32a2d000 - 0x32be7fff CoreGraphics armv6 <58e0bf5baaceb205917d03050a0ac954> /System/Library/Frameworks/CoreGraphics.framework/CoreGraphics
0x32c71000 - 0x32cedfff AddressBookUI armv6 <f617c0012f4d5caf1c0cc0fad2fc8c2e> /System/Library/Frameworks/AddressBookUI.framework/AddressBookUI
0x32cee000 - 0x32d58fff libstdc++.6.dylib armv6 <1982380fa8b60e63fd29559efb58d1f0> /usr/lib/libstdc++.6.dylib
0x32d59000 - 0x32d5afff IOMobileFramebuffer armv6 <436a03767758117cc9ecec8417f2d185> /System/Library/PrivateFrameworks/IOMobileFramebuffer.framework/IOMobileFramebuffer
0x32eb0000 - 0x32eb7fff ProtocolBuffer armv6 <faaee75832afeaf9d352318edd62970b> /System/Library/PrivateFrameworks/ProtocolBuffer.framework/ProtocolBuffer
0x32eb8000 - 0x32ef2fff CoreTelephony armv6 <56a514c096120e8a1e1bb52b8645699b> /System/Library/PrivateFrameworks/CoreTelephony.framework/CoreTelephony
0x32f11000 - 0x32f94fff PhotoLibrary armv6 <e37c2ebe58a035ef13164d834fa12699> /System/Library/PrivateFrameworks/PhotoLibrary.framework/PhotoLibrary
0x330ce000 - 0x330f1fff libRIP.A.dylib armv6 <5e02a84096d714b4f8a38aa8f4107b85> /System/Library/Frameworks/CoreGraphics.framework/Resources/libRIP.A.dylib
0x33110000 - 0x33148fff TextInput armv6 <a13cc5ea542bf5d4a66256e1c7ca0cee> /System/Library/PrivateFrameworks/TextInput.framework/TextInput
0x3314b000 - 0x33f3cfff UIKit armv6 <dfcffbbb1e8813d91fe5fd5daf69ed11> /System/Library/Frameworks/UIKit.framework/UIKit
0x33f74000 - 0x33fd5fff GMM armv6 <45ce64a42afbb59806bf588439b4bfc6> /System/Library/PrivateFrameworks/GMM.framework/GMM
0x33fe5000 - 0x34040fff CoreAudio armv6 <92290f4dba14837427eaa399a55929cc> /System/Library/Frameworks/CoreAudio.framework/CoreAudio
0x34150000 - 0x34156fff MBX2D armv6 <0e5d686d0e641a8efc5dc2003c37e2f7> /System/Library/PrivateFrameworks/MBX2D.framework/MBX2D
0x34170000 - 0x3418efff AddressBook armv6 <d221b21a3d8594abc8ee717ae6f77076> /System/Library/Frameworks/AddressBook.framework/AddressBook
0x3419b000 - 0x341a5fff MobileCoreServices armv6 <48254d8d389dc9ecf90688e227ddba1a> /System/Library/Frameworks/MobileCoreServices.framework/MobileCoreServices
0x341f0000 - 0x341f4fff IOSurface armv6 <1b951baaae45efbbae825b670612f478> /System/Library/PrivateFrameworks/IOSurface.framework/IOSurface
0x341f5000 - 0x342f6fff libicucore.A.dylib armv6 <aceb70042ae76cfec9838f6771db48d8> /usr/lib/libicucore.A.dylib
0x3430a000 - 0x343adfff libobjc.A.dylib armv6 <c3df9efc40404bc1865caacbf6209a1e> /usr/lib/libobjc.A.dylib

Here's a crazy thought: install 3.1 on your iPhone so you can test your app and keep your client happy. Exactly how long would it take you to install 3.1?

The first log, there's just not enough info in the stack to say why it crashed. It seems most likely that some UI element was released somewhere and still being referenced, though that is kind of a wild guess.
The second one though, is telling you that when the user pressed the lock button while in your app, it took too long to return from the method that processed the notification the application was suspending - so it was killed. You need to check whatever method you have activating on suspend and see what is taking so long to run there.