我有一个数据集，其中样本数为 25000，特征数为 24995。我正在尝试在此数据上训练 keras 自动编码器模型并面临 OOM 错误。该模型的一些细节是
Input matrix shape : (25000, 24995)
此输入矩阵分为验证集，分别作为训练数据和测试数据。
Train Matrix shape : (18750, 24995)
Test Matrix shape : (6250, 24995)
训练代码是
from keras.layers import Input, Dense
input_layer = Input(shape=(train_matrix.shape[1],))
encoding_hlayer1_dims = 12500
encoding_hlayer1 = Dense(encoding_hlayer1_dims, activation='relu', trainable=True, name="layer1")(input_layer)
decoding_hlayer1 = Dense(train_matrix.shape[1], activation='relu')(encoding_hlayer1)
autoencoder = Model(input_layer, decoding_hlayer1)
autoencoder.compile(optimizer='adam', loss='binary_crossentropy')
模型的摘要为
Layer (type) Output Shape Param #
=================================================================
input_2 (InputLayer) (None, 24995) 0
_________________________________________________________________
layer1 (Dense) (None, 12500) 312450000
_________________________________________________________________
dense_1 (Dense) (None, 24995) 312462495
=================================================================
Total params: 624,912,495
Trainable params: 624,912,495
Non-trainable params: 0
用于训练模型的代码
## Train
history = autoencoder.fit(train_matrix.toarray(), train_matrix.toarray(),
epochs=50,
batch_size=64,
shuffle=True,
validation_data=(test_matrix.toarray(), test_matrix.toarray()))
当我开始训练模式时，出现以下错误：
ResourceExhaustedError (see above for traceback): OOM when allocating tensor with shape[24995,12500]
[[Node: mul_3 = Mul[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/gpu:0"](beta_1/read, Variable/read)]]
我正在使用 2 个 Nvidia Tesla K40c GPU，每个 12G。据我所知，该模型应适合内存，为 25000 * 12500 * 2 = 0.625 GB。此外，输入矩阵 dtype 是 numpy.float32。
谁能指出我在这里到底做错了什么？
更新：完整的错误日志
Train on 18750 samples, validate on 6250 samples
Epoch 1/100
ResourceExhaustedErrorTraceback (most recent call last)
<ipython-input-8-503b20168fa5> in <module>()
6 batch_size=4096,
7 shuffle=True,
----> 8 validation_data=(test_matrix.toarray(), test_matrix.toarray()))
9 # autoencoder.save("/tmp/Models/sae_models/epochs_" + str(epochs) + ".model", include_optimizer=True)
10
/usr/local/lib/python2.7/dist-packages/keras/engine/training.pyc in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, **kwargs)
1428 val_f=val_f, val_ins=val_ins, shuffle=shuffle,
1429 callback_metrics=callback_metrics,
-> 1430 initial_epoch=initial_epoch)
1431
1432 def evaluate(self, x, y, batch_size=32, verbose=1, sample_weight=None):
/usr/local/lib/python2.7/dist-packages/keras/engine/training.pyc in _fit_loop(self, f, ins, out_labels, batch_size, epochs, verbose, callbacks, val_f, val_ins, shuffle, callback_metrics, initial_epoch)
1077 batch_logs['size'] = len(batch_ids)
1078 callbacks.on_batch_begin(batch_index, batch_logs)
-> 1079 outs = f(ins_batch)
1080 if not isinstance(outs, list):
1081 outs = [outs]
/usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.pyc in __call__(self, inputs)
2263 value = (indices, sparse_coo.data, sparse_coo.shape)
2264 feed_dict[tensor] = value
-> 2265 session = get_session()
2266 updated = session.run(self.outputs + [self.updates_op],
2267 feed_dict=feed_dict,
/usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.pyc in get_session()
166 if not _MANUAL_VAR_INIT:
167 with session.graph.as_default():
--> 168 _initialize_variables()
169 return session
170
/usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.pyc in _initialize_variables()
339 if uninitialized_variables:
340 sess = get_session()
--> 341 sess.run(tf.variables_initializer(uninitialized_variables))
342
343
/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in run(self, fetches, feed_dict, options, run_metadata)
787 try:
788 result = self._run(None, fetches, feed_dict, options_ptr,
--> 789 run_metadata_ptr)
790 if run_metadata:
791 proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)
/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _run(self, handle, fetches, feed_dict, options, run_metadata)
995 if final_fetches or final_targets:
996 results = self._do_run(handle, final_targets, final_fetches,
--> 997 feed_dict_string, options, run_metadata)
998 else:
999 results = []
/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _do_run(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)
1130 if handle is None:
1131 return self._do_call(_run_fn, self._session, feed_dict, fetch_list,
-> 1132 target_list, options, run_metadata)
1133 else:
1134 return self._do_call(_prun_fn, self._session, handle, feed_dict,
/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _do_call(self, fn, *args)
1150 except KeyError:
1151 pass
-> 1152 raise type(e)(node_def, op, message)
1153
1154 def _extend_graph(self):
ResourceExhaustedError: OOM when allocating tensor with shape[24995,12500]
[[Node: layer1/kernel/Assign = Assign[T=DT_FLOAT, _class=["loc:@layer1/kernel"], use_locking=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/gpu:0"](layer1/kernel, layer1/random_uniform)]]
Caused by op u'layer1/kernel/Assign', defined at:
File "/usr/lib/python2.7/runpy.py", line 174, in _run_module_as_main
"__main__", fname, loader, pkg_name)
File "/usr/lib/python2.7/runpy.py", line 72, in _run_code
exec code in run_globals
File "/usr/local/lib/python2.7/dist-packages/ipykernel_launcher.py", line 16, in <module>
app.launch_new_instance()
File "/usr/local/lib/python2.7/dist-packages/traitlets/config/application.py", line 658, in launch_instance
app.start()
File "/usr/local/lib/python2.7/dist-packages/ipykernel/kernelapp.py", line 477, in start
ioloop.IOLoop.instance().start()
File "/usr/local/lib/python2.7/dist-packages/zmq/eventloop/ioloop.py", line 177, in start
super(ZMQIOLoop, self).start()
File "/usr/local/lib/python2.7/dist-packages/tornado/ioloop.py", line 888, in start
handler_func(fd_obj, events)
File "/usr/local/lib/python2.7/dist-packages/tornado/stack_context.py", line 277, in null_wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/zmq/eventloop/zmqstream.py", line 440, in _handle_events
self._handle_recv()
File "/usr/local/lib/python2.7/dist-packages/zmq/eventloop/zmqstream.py", line 472, in _handle_recv
self._run_callback(callback, msg)
File "/usr/local/lib/python2.7/dist-packages/zmq/eventloop/zmqstream.py", line 414, in _run_callback
callback(*args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/tornado/stack_context.py", line 277, in null_wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/ipykernel/kernelbase.py", line 283, in dispatcher
return self.dispatch_shell(stream, msg)
File "/usr/local/lib/python2.7/dist-packages/ipykernel/kernelbase.py", line 235, in dispatch_shell
handler(stream, idents, msg)
File "/usr/local/lib/python2.7/dist-packages/ipykernel/kernelbase.py", line 399, in execute_request
user_expressions, allow_stdin)
File "/usr/local/lib/python2.7/dist-packages/ipykernel/ipkernel.py", line 196, in do_execute
res = shell.run_cell(code, store_history=store_history, silent=silent)
File "/usr/local/lib/python2.7/dist-packages/ipykernel/zmqshell.py", line 533, in run_cell
return super(ZMQInteractiveShell, self).run_cell(*args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/IPython/core/interactiveshell.py", line 2718, in run_cell
interactivity=interactivity, compiler=compiler, result=result)
File "/usr/local/lib/python2.7/dist-packages/IPython/core/interactiveshell.py", line 2822, in run_ast_nodes
if self.run_code(code, result):
File "/usr/local/lib/python2.7/dist-packages/IPython/core/interactiveshell.py", line 2882, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "<ipython-input-4-ee2fe8e92d7c>", line 4, in <module>
encoding_hlayer1 = Dense(encoding_hlayer1_dims, activation='relu', trainable=True, name="layer1")(input_layer)
File "/usr/local/lib/python2.7/dist-packages/keras/engine/topology.py", line 569, in __call__
self.build(input_shapes[0])
File "/usr/local/lib/python2.7/dist-packages/keras/layers/core.py", line 825, in build
constraint=self.kernel_constraint)
File "/usr/local/lib/python2.7/dist-packages/keras/legacy/interfaces.py", line 87, in wrapper
return func(*args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/keras/engine/topology.py", line 391, in add_weight
weight = K.variable(initializer(shape), dtype=dtype, name=name)
File "/usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.py", line 321, in variable
v = tf.Variable(value, dtype=_convert_string_dtype(dtype), name=name)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variables.py", line 200, in __init__
expected_shape=expected_shape)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variables.py", line 309, in _init_from_args
validate_shape=validate_shape).op
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/state_ops.py", line 271, in assign
validate_shape=validate_shape)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gen_state_ops.py", line 45, in assign
use_locking=use_locking, name=name)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op
op_def=op_def)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2506, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 1269, in __init__
self._traceback = _extract_stack()
ResourceExhaustedError (see above for traceback): OOM when allocating tensor with shape[24995,12500]
[[Node: layer1/kernel/Assign = Assign[T=DT_FLOAT, _class=["loc:@layer1/kernel"], use_locking=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/gpu:0"](layer1/kernel, layer1/random_uniform)]]
参数的总数根据您的代码624,912,495。这应该需要624912495 * 4 / 1024**3 = 2.32GB 来存储权重(而不是您计算的 0.625(。
除此之外，您还需要为优化器存储初始值设定项和至少另外 3 个副本，一个分别用于一阶和二阶动量，一个用于实际更新，更不用说一些临时存储计算，因为任何时候你写a + b，你需要内存来存储它，并且可能有一些隐藏的。
总体而言，您很快就会发现总内存使用量远高于 12 GB，这就是内存不足的原因。
您可以尝试使用使用较少内存的 SGD 优化器，但您仍然可能会用完。