1. 技术文章
  2. 【tf.keras】Resource exhausted: OOM when allocating tensor with shape [9216,4096] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc

【tf.keras】Resource exhausted: OOM when allocating tensor with shape [9216,4096] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc

由 亡梦爱人 提交于 2020-01-18 07:48:54

运行以下类似代码：

while True:
    inputs, outputs = get_AlexNet()
    model = tf.keras.Model(inputs=inputs, outputs=outputs)
    
    model.summary()

    adam_opt = tf.keras.optimizers.Adam(learning_rate)
    # The compile step specifies the training configuration.
    model.compile(optimizer=adam_opt, loss='categorical_crossentropy', metrics=['accuracy'])

    # load weights from h5 file
    model.load_weights('alexnet_weights.h5')

最后会报错：

OP_REQUIRES failed at cwise_ops_common.cc:70 : Resource exhausted: OOM when allocating tensor with shape[9216,4096] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc

解决办法：

from keras import backend as K
K.clear_session()

如：

from keras import backend as K

while True:
    # 清空之前model占用的内存，防止OOM
    K.clear_session()

    inputs, outputs = get_AlexNet()
    model = tf.keras.Model(inputs=inputs, outputs=outputs)
    
    model.summary()

    adam_opt = tf.keras.optimizers.Adam(learning_rate)
    # The compile step specifies the training configuration.
    model.compile(optimizer=adam_opt, loss='categorical_crossentropy', metrics=['accuracy'])

    # load weights from h5 file
    model.load_weights('alexnet_weights.h5')

详细报错如下：

2019-06-03 21:54:24.789150: W T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:275] Allocator (GPU_0_bfc) ran out of memory trying to allocate 144.00MiB.  Current allocation summary follows.
2019-06-03 21:54:24.804684: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (256):  Total Chunks: 243, Chunks in use: 243. 60.8KiB allocated for chunks. 60.8KiB in use in bin. 6.6KiB client-requested in use in bin.
2019-06-03 21:54:24.813190: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (512):  Total Chunks: 19, Chunks in use: 19. 14.3KiB allocated for chunks. 14.3KiB in use in bin. 14.3KiB client-requested in use in bin.
2019-06-03 21:54:24.841197: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (1024):     Total Chunks: 52, Chunks in use: 52. 62.5KiB allocated for chunks. 62.5KiB in use in bin. 60.6KiB client-requested in use in bin.
2019-06-03 21:54:24.843308: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (2048):     Total Chunks: 2, Chunks in use: 2. 5.0KiB allocated for chunks. 5.0KiB in use in bin. 3.0KiB client-requested in use in bin.
2019-06-03 21:54:24.844847: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (4096):     Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.846267: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (8192):     Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.848125: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (16384):    Total Chunks: 31, Chunks in use: 31. 511.0KiB allocated for chunks. 511.0KiB in use in bin. 496.0KiB client-requested in use in bin.
2019-06-03 21:54:24.849356: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (32768):    Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.850511: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (65536):    Total Chunks: 16, Chunks in use: 16. 1.43MiB allocated for chunks. 1.43MiB in use in bin. 1.42MiB client-requested in use in bin.
2019-06-03 21:54:24.852015: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (131072):   Total Chunks: 23, Chunks in use: 23. 3.72MiB allocated for chunks. 3.72MiB in use in bin. 3.46MiB client-requested in use in bin.
2019-06-03 21:54:24.863147: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (262144):   Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.864633: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (524288):   Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.865992: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (1048576):  Total Chunks: 17, Chunks in use: 17. 21.15MiB allocated for chunks. 21.15MiB in use in bin. 19.92MiB client-requested in use in bin.
2019-06-03 21:54:24.867384: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (2097152):  Total Chunks: 52, Chunks in use: 52. 144.75MiB allocated for chunks. 144.75MiB in use in bin. 137.86MiB client-requested in use in bin.
2019-06-03 21:54:24.868803: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (4194304):  Total Chunks: 3, Chunks in use: 3. 17.16MiB allocated for chunks. 17.16MiB in use in bin. 10.13MiB client-requested in use in bin.
2019-06-03 21:54:24.870144: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (8388608):  Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.871061: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (16777216):     Total Chunks: 3, Chunks in use: 2. 62.97MiB allocated for chunks. 42.20MiB in use in bin. 37.19MiB client-requested in use in bin.
2019-06-03 21:54:24.871849: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (33554432):     Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.874994: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (67108864):     Total Chunks: 21, Chunks in use: 21. 1.40GiB allocated for chunks. 1.40GiB in use in bin. 1.31GiB client-requested in use in bin.
2019-06-03 21:54:24.875718: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (134217728):    Total Chunks: 20, Chunks in use: 20. 2.98GiB allocated for chunks. 2.98GiB in use in bin. 2.81GiB client-requested in use in bin.
2019-06-03 21:54:24.876800: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:630] Bin (268435456):    Total Chunks: 0, Chunks in use: 0. 0B allocated for chunks. 0B in use in bin. 0B client-requested in use in bin.
2019-06-03 21:54:24.877455: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:646] Bin for 144.00MiB was 128.00MiB, Chunk State: 
2019-06-03 21:54:24.877906: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:665] Chunk at 0000000B03E00000 of size 1280
2019-06-03 21:54:24.878316: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:665] Chunk at 0000000B03E00500 of size 256
2019-06-03 21:54:24.879415: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:665] Chunk at 0000000B03E00600 of size 256
2019-06-03 21:54:24.879816: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:665] Chunk at 0000000B03E00700 of size 256
...
2019-06-03 21:54:24.998647: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:674] 1 Chunks of size 256733696 totalling 244.84MiB
2019-06-03 21:54:24.998857: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:678] Sum Total of in-use chunks: 4.60GiB
2019-06-03 21:54:24.999076: I T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:680] Stats: 
Limit:                  4965636505
InUse:                  4943860224
MaxInUse:               4943860224
NumAllocs:                 2362778
MaxAllocSize:            516972544

2019-06-03 21:54:24.999520: W T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:279] ********x************************************************************************x*****************x
2019-06-03 21:54:25.001526: W T:\src\github\tensorflow\tensorflow\core\framework\op_kernel.cc:1275] OP_REQUIRES failed at cwise_ops_common.cc:70 : Resource exhausted: OOM when allocating tensor with shape[9216,4096] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
2019-06-03 21:54:25.108672: W T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 372.96MiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-06-03 21:54:25.129713: W T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 482.40MiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2019-06-03 21:54:25.145367: W T:\src\github\tensorflow\tensorflow\core\common_runtime\bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 331.52MiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
Traceback (most recent call last):
  File "E:/PycharmProjects/ActiveLearning/AlexNet_AL.py", line 156, in <module>
    validation_data=(x_val, y_val))
  File "C:\taotao\Python\Python36\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1363, in fit
    validation_steps=validation_steps)
  File "C:\taotao\Python\Python36\lib\site-packages\tensorflow\python\keras\engine\training_arrays.py", line 264, in fit_loop
    outs = f(ins_batch)
  File "C:\taotao\Python\Python36\lib\site-packages\tensorflow\python\keras\backend.py", line 2914, in __call__
    fetched = self._callable_fn(*array_vals)
  File "C:\taotao\Python\Python36\lib\site-packages\tensorflow\python\client\session.py", line 1382, in __call__
    run_metadata_ptr)
  File "C:\taotao\Python\Python36\lib\site-packages\tensorflow\python\framework\errors_impl.py", line 519, in __exit__
    c_api.TF_GetCode(self.status.status))
tensorflow.python.framework.errors_impl.ResourceExhaustedError: OOM when allocating tensor with shape[9216,4096] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
     [[Node: training_4/Adam/gradients/dense/kernel/Regularizer/Square_grad/Mul_1 = Mul[T=DT_FLOAT, _class=["loc:@training_4/Adam/gradients/AddN_5"], _device="/job:localhost/replica:0/task:0/device:GPU:0"](dense/kernel/Regularizer/Square/ReadVariableOp, training_4/Adam/gradients/dense/kernel/Regularizer/Square_grad/Mul)]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

     [[Node: metrics_4/acc/Mean/_1023 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device_incarnation=1, tensor_name="edge_1423_metrics_4/acc/Mean", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"]()]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

References

Keras解决OOM超内存问题 -- silent56_th
Keras 循环训练模型跑数据时内存泄漏的问题解决办法 -- jemmie_w

标签
tensorflow
runtime
bfc
python
易学教程内所有资源均来自网络或用户发布的内容，如有违反法律规定的内容欢迎反馈！
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!