search

carpedm20 / lstm-char-cnn-tensorflow

in progress
MIT License
769 stars 243 forks source link

did anybody successfully train? #5

Closed badend closed 8 years ago

badend commented 8 years ago

with GPU I just reduced a lot of options such as batch size, feature_maps etc.

Always OOM occured!!

Resource exhausted: OOM when allocating tensor with shape[50,450537] W tensorflow/core/common_runtime/executor.cc:1102] 0x8bcfce0 Compute status: Resource exhausted: OOM when allocating tensor with shape[50,450537] [[Node: LSTMTDNN/LSTM/Linear_34/MatMul = MatMul[T=DT_FLOAT, transpose_a=false, transpose_b=false, _device="/job:localhost/replica:0/task:0/gpu:0"](LSTMTDNN/LSTM/dropout_34/mul_1, LSTMTDNN/LSTM/Linear/Matrix/read)]] I tensorflow/core/common_runtime/gpu/pool_allocator.cc:244] PoolAllocator: After 2987 get requests, put_count=1597 evicted_count=1000 eviction_rate=0.626174 and unsatisfied allocation rate=0.833612

carpedm20 commented 8 years ago

@badend Resource exhausted means your GPU doesn't have enough memory to build a graph which is needed for this repo. You'd better reduce the size of embedding or # of kernels.