【自己幫影片換主角】FakeApp 教學
挽歌之聲
971 回覆
546 Like
50 Dislike
前日做左一段覺得少少似.唔小心del folder del 埋
唔係時間耐就一定會低落去
要睇番夠唔夠素材比佢train，唔夾就唔夾
train到咁上下就會停左
要睇番夠唔夠素材比佢train，唔夾就唔夾
train到咁上下就會停左
素材真係一個超大問題
訓一晚等佢LOAD算
ResourceExhaustedError (see above for traceback): OOM when allocating tensor wit
h shape[2,512,9,9]
[[Node: model_1/conv2d_4/convolution = Conv2D[T=DT_FLOAT, data_format="
NHWC", padding="SAME", strides=[1, 2, 2, 1], use_cudnn_on_gpu=true, _device="/jo
b:localhost/replica:0/task:0/device:GPU:0"](model_1/leaky_re_lu_3/sub, conv2d_4/
kernel/read)]]
[[Node: loss_1/mul/_351 = _Recv[client_terminated=false, recv_device="/
job:localhost/replica:0/task:0/device:CPU:0", send_device="/job:localhost/replic
a:0/task:0/device:GPU:0", send_device_incarnation=1, tensor_name="edge_1601_loss
_1/mul", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:C
PU:0"]()]]
[5960] Failed to execute script train
咩事
h shape[2,512,9,9]
[[Node: model_1/conv2d_4/convolution = Conv2D[T=DT_FLOAT, data_format="
NHWC", padding="SAME", strides=[1, 2, 2, 1], use_cudnn_on_gpu=true, _device="/jo
b:localhost/replica:0/task:0/device:GPU:0"](model_1/leaky_re_lu_3/sub, conv2d_4/
kernel/read)]]
[[Node: loss_1/mul/_351 = _Recv[client_terminated=false, recv_device="/
job:localhost/replica:0/task:0/device:CPU:0", send_device="/job:localhost/replic
a:0/task:0/device:GPU:0", send_device_incarnation=1, tensor_name="edge_1601_loss
_1/mul", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:C
PU:0"]()]]
[5960] Failed to execute script train
咩事
你用咩卡? 我1080ti load左二十個hr都係0.017
It's out of memory, try lowering layers or nodes.
reddit話係咁
reddit話係咁
Lm
run緊吉澤同stephy
想知整咗一次model之後del咗data b舊圖（結衣bb)換第2d圖(結衣bb 其他相）嘅話
喺加新嘢落model定當重新整過？
喺加新嘢落model定當重新整過？
咁你A係咪同一個先
宜家喺同一個，因為b嘅data好似未夠，想加多d相再train好啲
但遲下a會換第2條影片
但遲下a會換第2條影片
ResourceExhaustedError (see above for traceback): OOM when allocating tensor with shape[5,5,512,1024]
[[Node: training/Adam/Variable_6/Assign = Assign[T=DT_FLOAT, _class=["loc:@training/Adam/Variable_6"], use_lock
ing=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/device:GPU:0"](training/Adam/Variable_6, trainin
g/Adam/Const_8)]]
[5388] Failed to execute script train
次次去到TRAIN都係咁
樓主救求
[[Node: training/Adam/Variable_6/Assign = Assign[T=DT_FLOAT, _class=["loc:@training/Adam/Variable_6"], use_lock
ing=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/device:GPU:0"](training/Adam/Variable_6, trainin
g/Adam/Const_8)]]
[5388] Failed to execute script train
次次去到TRAIN都係咁
樓主救求理論上喺咪只要b喺同一個人都train得好快？
2018-01-29 19:52:26.762355: W C:\tf_jenkins\home\workspace\rel-win\M\windows-gpu
\PY\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc
) ran out of memory trying to allocate 1.13GiB. The caller indicates that this i
s not a failure, but may mean that there could be performance gains if more memo
ry is available.
Loss: 0.162298 0.13392
Printing config file to c:\fakes\model\config.p
Saving model weights
Traceback (most recent call last):
File "train.py", line 138, in <module>
ValueError: cannot reshape array of size 294912 into shape (4,7,3,64,64,3)
[788] Failed to execute script train
又有新問題
\PY\35\tensorflow\core\common_runtime\bfc_allocator.cc:217] Allocator (GPU_0_bfc
) ran out of memory trying to allocate 1.13GiB. The caller indicates that this i
s not a failure, but may mean that there could be performance gains if more memo
ry is available.
Loss: 0.162298 0.13392
Printing config file to c:\fakes\model\config.p
Saving model weights
Traceback (most recent call last):
File "train.py", line 138, in <module>
ValueError: cannot reshape array of size 294912 into shape (4,7,3,64,64,3)
[788] Failed to execute script train
又有新問題
留名睇杜小喬
咁冇問題 可以繼續 唔刪model就得
就咁加相去data b定喺del咗data b舊相先？
depends 如果本身load到好低 咁可以刪
其實刪唔刪都得
其實刪唔刪都得