Python 在使用tensorflow库培训深度学习模型时，我遇到了一个错误：gpu上的ResourceExhausterRoom（128 gb RAM）请帮助我

C:\Users\CVL Acoustics\Documents\bangla句子校正母版>python train.py 坐下来放松一下，训练模特需要一些时间。。。 词汇量250000 警告：tensorflow:From C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\ops\rnn.py:417:使用seq\u dim调用反向序列（来自tensorflow.python.ops.array\u ops）已被弃用，并将在未来版本中删除。 更新说明： seq_dim已弃用，请改用seq_轴 警告：tensorflow:From C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\util\deprecation.py:432:使用批处理\u dim调用反向\u序列（来自tensorflow.python.ops.array\u ops）已被弃用，并将在未来版本中删除。 更新说明： 不推荐使用batch\u dim，请改用batch\u axis 警告：tensorflow:From train.py:228:softmax_cross_entropy_with_logits（来自tensorflow.python.ops.nn_ops）已弃用，将在未来版本中删除。 更新说明：

TensorFlow的未来主要版本将允许渐变流动 默认情况下，输入backprop上的标签

参见{tf.nn.softmax_cross_entropy_与_logits_v2}

第一纪元 培训回溯（最近一次呼叫最后一次）： 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\client\session.py”，第1322行，在调用中 返回fn（*args） 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\client\session.py”，第1307行，在\u run\u fn中 选项、提要、获取列表、目标列表、运行元数据） 文件“C:\Users\CVL austics\Anaconda3\lib\site packages\tensorflow\python\client\session.py”，第1409行，位于调用会话运行中 运行（元数据） tensorflow.python.framework.errors\u impl.ResourceExhaustedError:OOM当使用形状[6656250000]和类型float on/job:localhost/replica:0/task:0/device:GPU:0分配程序GPU\u bfc分配tensor时 [[Node:MatMul=MatMul[T=DT_FLOAT，transpose_a=false，transpose_b=false，_device=“/job:localhost/replica:0/task:0/device:GPU:0”]（重塑，变量_1/read）]] 提示：如果您想在OOM发生时查看已分配的张量列表，请在OOM上添加report_tensor_allocations_on_to RunOptions以获取当前分配信息

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

提示：如果您想在OOM发生时查看已分配的张量列表，请在OOM上添加report_tensor_allocations_on_to RunOptions以获取当前分配信息

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

在处理上述异常期间，发生了另一个异常：

回溯（最近一次呼叫最后一次）： 文件“train.py”，第321行，在 _，l=安全运行（[列车运行，损失]，fd） 文件“C:\Users\CVL austics\Anaconda3\lib\site packages\tensorflow\python\client\session.py”，第900行，正在运行 运行_元数据_ptr） 文件“C:\Users\CVL austics\Anaconda3\lib\site packages\tensorflow\python\client\session.py”，第1135行，正在运行 feed_dict_tensor、options、run_元数据） 文件“C:\Users\CVL austics\Anaconda3\lib\site packages\tensorflow\python\client\session.py”，第1316行，在运行中 运行（元数据） 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\client\session.py”，第1335行，在调用中 提升类型（e）（节点定义、操作、消息） tensorflow.python.framework.errors\u impl.ResourceExhaustedError:OOM当使用形状[6656250000]和类型float on/job:localhost/replica:0/task:0/device:GPU:0分配程序GPU\u bfc分配tensor时 [[Node:MatMul=MatMul[T=DT_FLOAT，transpose_a=false，transpose_b=false，_device=“/job:localhost/replica:0/task:0/device:GPU:0”]（重塑，变量_1/read）]] 提示：如果您想在OOM发生时查看已分配的张量列表，请在OOM上添加report_tensor_allocations_on_to RunOptions以获取当前分配信息

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

提示：如果您想在OOM发生时查看已分配的张量列表，请在OOM上添加report_tensor_allocations_on_to RunOptions以获取当前分配信息

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

由op“MatMul”引起，定义为： 文件“train.py”，第218行，在 解码器逻辑平坦=tf.add（tf.matmul（解码器输出平坦，W），b） 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\ops\math\u ops.py”，第2014行，matmul格式 a、 b，转置a=转置a，转置b=转置b，名称=名称） 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\ops\gen\u math\u ops.py”，第4278行，mat\u mul 名称=名称） 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\framework\op_def_library.py”，第787行，位于_apply_op_helper中 op_def=op_def） 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\framework\ops.py”，第3414行，位于create\u op op_def=op_def） 文件“C:\Users\CVL Acoustics\Anaconda3\lib\site packages\tensorflow\python\framework\ops.py”，第1740行，位于init self._traceback=self._graph._extract_stack（）35; pylint:disable=protected access

ResourceExhustederRor（回溯请参见上文）：当使用形状[6656250000]和类型float on/job:localhost/replica:0/任务：0/设备：GPU:0由分配器GPU\U 0\bfc分配张量时，OOM [[Node:MatMul=MatMul[T=DT_FLOAT，transpose_a=false，transpose_b=false，_device=“/job:localhost/replica:0/task:0/device:GPU:0”]（重塑，变量_1/read）]] 提示：如果您想在OOM发生时查看已分配的张量列表，请在OOM上添加report_tensor_allocations_on_to RunOptions以获取当前分配信息

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

---

提示：如果要在OOM发生时查看已分配的张量列表，请将“报告”张量“分配”添加到当前分配信息的RunOptions中。

出现这种情况的原因有几个

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

     [[Node: rnn/while/cond/Add/_87 = _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_421_rnn/while/cond/Add", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](^_clooprnn/while/cond/ArgMax/dimension/_1)]]

• 尝试减少网络的参数
• 尝试减少批处理大小
• 检查当前是否有另一个正在分配内存的内核处于活动状态

---

听起来好像内存不足。您能帮我解决这个问题吗@JammyD