我正在尝试在张量流中创建自定义损失函数。我正在使用tensorflow v2.0.rc0运行代码。以下是代码，函数min_dist_loss计算神经网络输出之间的成对损耗。这是代码
def min_dist_loss(_, y_pred):
distances = []
for i in range(0, 16):
for j in range(i + 1, 16):
distances.append(tf.linalg.norm(y_pred[i] - y_pred[j]))
return -tf.reduce_min(distances)
并且模块正在按以下方式初始化和编译
import tensorflow as tf
from tensorboard.plugins.hparams import api as hp
HP_NUM_UNITS = hp.HParam('num_units', hp.Discrete([6, 7]))
HP_OPTIMIZER = hp.HParam('optimizer', hp.Discrete(['adam', 'sgd']))
METRIC_ACCURACY = 'accuracy'
with tf.summary.create_file_writer('logs\hparam_tuning').as_default():
hp.hparams_config(
hparams=[HP_NUM_UNITS, HP_OPTIMIZER],
metrics=[hp.Metric(METRIC_ACCURACY, display_name='Accuracy')]
)
def train_test_model(logdir, hparams):
weight1 = np.random.normal(loc=0.0, scale=0.01, size=[4, hparams[HP_NUM_UNITS]])
init1 = tf.constant_initializer(weight1)
weight2 = np.random.normal(loc=0.0, scale=0.01, size=[hparams[HP_NUM_UNITS], 7])
init2 = tf.constant_initializer(weight2)
model = tf.keras.models.Sequential([
# tf.keras.layers.Flatten(),
tf.keras.layers.Dense(hparams[HP_NUM_UNITS], activation=tf.nn.sigmoid, kernel_initializer=init1),
tf.keras.layers.Dense(7, activation=tf.nn.sigmoid, kernel_initializer=init2) if hparams[HP_NUM_UNITS] == 6 else
None,
])
model.compile(
optimizer=hparams[HP_OPTIMIZER],
loss=min_dist_loss,
# metrics=['accuracy'],
)
x_train = [list(k) for k in itertools.product([0, 1], repeat=4)]
shuffle(x_train)
x_train = 2 * np.array(x_train) - 1
model.fit(
x_train, epochs=1, batch_size=16,
callbacks=[
tf.keras.callbacks.TensorBoard(logdir),
hp.KerasCallback(logdir, hparams)
],
)
现在，由于y_pred中的张量对象min_dist_loss是形状为[?, 7]的对象，因此用i进行索引将引发以下错误：
Traceback (most recent call last):
File "/home/pc/Documents/user/code/keras_tensorflow/src/try1.py", line 95, in <module>
run('logs\hparam_tuning' + run_name, hparams)
File "/home/pc/Documents/user/code/keras_tensorflow/src/try1.py", line 78, in run
accuracy = train_test_model(run_dir, hparams)
File "/home/pc/Documents/user/code/keras_tensorflow/src/try1.py", line 66, in train_test_model
hp.KerasCallback(logdir, hparams)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training.py", line 734, in fit
use_multiprocessing=use_multiprocessing)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_v2.py", line 324, in fit
total_epochs=epochs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_v2.py", line 123, in run_one_epoch
batch_outs = execution_function(iterator)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_v2_utils.py", line 86, in execution_function
distributed_function(input_fn))
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/eager/def_function.py", line 427, in __call__
self._initialize(args, kwds, add_initializers_to=initializer_map)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/eager/def_function.py", line 370, in _initialize
*args, **kwds))
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/eager/function.py", line 1847, in _get_concrete_function_internal_garbage_collected
graph_function, _, _ = self._maybe_define_function(args, kwargs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/eager/function.py", line 2147, in _maybe_define_function
graph_function = self._create_graph_function(args, kwargs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/eager/function.py", line 2038, in _create_graph_function
capture_by_value=self._capture_by_value),
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/framework/func_graph.py", line 915, in func_graph_from_py_func
func_outputs = python_func(*func_args, **func_kwargs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/eager/def_function.py", line 320, in wrapped_fn
return weak_wrapped_fn().__wrapped__(*args, **kwds)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_v2_utils.py", line 73, in distributed_function
per_replica_function, args=(model, x, y, sample_weights))
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/distribute/distribute_lib.py", line 760, in experimental_run_v2
return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/distribute/distribute_lib.py", line 1787, in call_for_each_replica
return self._call_for_each_replica(fn, args, kwargs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/distribute/distribute_lib.py", line 2132, in _call_for_each_replica
return fn(*args, **kwargs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/autograph/impl/api.py", line 292, in wrapper
return func(*args, **kwargs)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_v2_utils.py", line 264, in train_on_batch
output_loss_metrics=model._output_loss_metrics)
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_eager.py", line 311, in train_on_batch
output_loss_metrics=output_loss_metrics))
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_eager.py", line 252, in _process_single_batch
training=training))
File "/home/pc/Documents/user/code/keras_tensorflow/venv/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training_eager.py", line 166, in _model_loss
per_sample_losses = loss_fn.call(targets[i], outs[i])
IndexError: list index out of range
如何计算此设置中的最小距离？任何帮助表示赞赏。另外，如果代码的其他部分有任何错误，请随时指出。我不熟悉在keras上使用tensorflow。
答案 0 :(得分：1)
Keras希望您也提供真实的标签。由于您正在定义自己的损失函数，并且未使用真实标签，因此可以传递一些垃圾标签。例如：np.arange(16)。
如下更改您的model.fit，它应该可以工作
model.fit(
x_train, np.arange(x_train.shape[0]), epochs=1, batch_size=16,
callbacks=[
tf.keras.callbacks.TensorBoard(logdir),
hp.KerasCallback(logdir, hparams)
],
)