Hi, \n In settings I have enabled my GPU. \n I was trying to implement the CuDNNGRU keras on kaggle. This is my model,
\n\ninp = kl.Input(shape=(max_words,))\nm = kl.Embedding(input_dim = cl+1, output_dim = EMBEDDING_DIM, weights=[embedding_matrix], trainable = False)(inp)\n#del(embedding_matrix);\nm = kl.SpatialDropout1D(rate = 0.25)(m)\nm = kl.Bidirectional(kl.CuDNNGRU(128, kernel_initializer=glorot_normal(seed=1029), \n recurrent_initializer=orthogonal(gain=1.0, seed=1029), name = 'bi_gru'))(m)\nm = kl.Dense(32, activation=\"relu\")(m)\nm = kl.Dropout(0.2)(m)\nm = kl.Dense(1, activation=\"sigmoid\")(m)\nmodel = Model(inputs=inp, outputs=m)\n\nmodel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\nprint(model.summary())\n\nnum_epochs = 5; batchSize = 128;\nAfter this this is my output.
\n\n(33927, 300)\n_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\ninput_1 (InputLayer) (None, 52) 0 \n_________________________________________________________________\nembedding_1 (Embedding) (None, 52, 300) 10178100 \n_________________________________________________________________\nspatial_dropout1d_1 (Spatial (None, 52, 300) 0 \n_________________________________________________________________\nbidirectional_1 (Bidirection (None, 256) 330240 \n_________________________________________________________________\ndense_1 (Dense) (None, 32) 8224 \n_________________________________________________________________\ndropout_1 (Dropout) (None, 32) 0 \n_________________________________________________________________\ndense_2 (Dense) (None, 1) 33 \n=================================================================\nTotal params: 10,516,597\nTrainable params: 338,497\nNon-trainable params: 10,178,100\nBut when i fit my model, \nmodel.fit(x_train, y_train, validation_data=(x_val, y_val), batch_size = batch_size, epochs=num_epochs, verbose = True);
\n\nIt is showing this error,
\n\nTrain on 1044898 samples, validate on 261224 samples
\n\nInvalidArgumentError Traceback (most recent call last)\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _do_call(self, fn, *args)\n 1333 try:\n-> 1334 return fn(*args)\n 1335 except errors.OpError as e:
\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _run_fn(feed_dict, fetch_list, target_list, options, run_metadata)\n 1316 # Ensure any changes to the graph are reflected in the runtime.\n-> 1317 self._extend_graph()\n 1318 return self._call_tf_sessionrun(
\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _extend_graph(self)\n 1351 with self._graph._session_run_lock(): # pylint: disable=protected-access\n-> 1352 tf_session.ExtendSession(self._session)\n 1353
\n\nInvalidArgumentError: No OpKernel was registered to support Op 'CudnnRNN' with these attrs. Registered devices: [CPU,XLA_CPU], Registered kernels:
", "messages": [ { "id": 456661, "postDate": "2019-01-16T09:01:38.117Z", "content": "Hi, \n In settings I have enabled my GPU. \n I was trying to implement the CuDNNGRU keras on kaggle. This is my model,
\n\ninp = kl.Input(shape=(max_words,))\nm = kl.Embedding(input_dim = cl+1, output_dim = EMBEDDING_DIM, weights=[embedding_matrix], trainable = False)(inp)\n#del(embedding_matrix);\nm = kl.SpatialDropout1D(rate = 0.25)(m)\nm = kl.Bidirectional(kl.CuDNNGRU(128, kernel_initializer=glorot_normal(seed=1029), \n recurrent_initializer=orthogonal(gain=1.0, seed=1029), name = 'bi_gru'))(m)\nm = kl.Dense(32, activation=\"relu\")(m)\nm = kl.Dropout(0.2)(m)\nm = kl.Dense(1, activation=\"sigmoid\")(m)\nmodel = Model(inputs=inp, outputs=m)\n\nmodel.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\nprint(model.summary())\n\nnum_epochs = 5; batchSize = 128;\nAfter this this is my output.
\n\n(33927, 300)\n_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\ninput_1 (InputLayer) (None, 52) 0 \n_________________________________________________________________\nembedding_1 (Embedding) (None, 52, 300) 10178100 \n_________________________________________________________________\nspatial_dropout1d_1 (Spatial (None, 52, 300) 0 \n_________________________________________________________________\nbidirectional_1 (Bidirection (None, 256) 330240 \n_________________________________________________________________\ndense_1 (Dense) (None, 32) 8224 \n_________________________________________________________________\ndropout_1 (Dropout) (None, 32) 0 \n_________________________________________________________________\ndense_2 (Dense) (None, 1) 33 \n=================================================================\nTotal params: 10,516,597\nTrainable params: 338,497\nNon-trainable params: 10,178,100\nBut when i fit my model, \nmodel.fit(x_train, y_train, validation_data=(x_val, y_val), batch_size = batch_size, epochs=num_epochs, verbose = True);
\n\nIt is showing this error,
\n\nTrain on 1044898 samples, validate on 261224 samples
\n\nInvalidArgumentError Traceback (most recent call last)\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _do_call(self, fn, *args)\n 1333 try:\n-> 1334 return fn(*args)\n 1335 except errors.OpError as e:
\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _run_fn(feed_dict, fetch_list, target_list, options, run_metadata)\n 1316 # Ensure any changes to the graph are reflected in the runtime.\n-> 1317 self._extend_graph()\n 1318 return self._call_tf_sessionrun(
\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _extend_graph(self)\n 1351 with self._graph._session_run_lock(): # pylint: disable=protected-access\n-> 1352 tf_session.ExtendSession(self._session)\n 1353
\n\nInvalidArgumentError: No OpKernel was registered to support Op 'CudnnRNN' with these attrs. Registered devices: [CPU,XLA_CPU], Registered kernels:
", "rawMarkdown": "Hi, \n In settings I have enabled my GPU. \n I was trying to implement the CuDNNGRU keras on kaggle. This is my model,\n\n inp = kl.Input(shape=(max_words,))\n m = kl.Embedding(input_dim = cl+1, output_dim = EMBEDDING_DIM, weights=[embedding_matrix], trainable = False)(inp)\n #del(embedding_matrix);\n m = kl.SpatialDropout1D(rate = 0.25)(m)\n m = kl.Bidirectional(kl.CuDNNGRU(128, kernel_initializer=glorot_normal(seed=1029), \n recurrent_initializer=orthogonal(gain=1.0, seed=1029), name = 'bi_gru'))(m)\n m = kl.Dense(32, activation=\"relu\")(m)\n m = kl.Dropout(0.2)(m)\n m = kl.Dense(1, activation=\"sigmoid\")(m)\n model = Model(inputs=inp, outputs=m)\n \n model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n print(model.summary())\n \n num_epochs = 5; batchSize = 128;\n\nAfter this this is my output.\n\n (33927, 300)\n _________________________________________________________________\n Layer (type) Output Shape Param # \n =================================================================\n input_1 (InputLayer) (None, 52) 0 \n _________________________________________________________________\n embedding_1 (Embedding) (None, 52, 300) 10178100 \n _________________________________________________________________\n spatial_dropout1d_1 (Spatial (None, 52, 300) 0 \n _________________________________________________________________\n bidirectional_1 (Bidirection (None, 256) 330240 \n _________________________________________________________________\n dense_1 (Dense) (None, 32) 8224 \n _________________________________________________________________\n dropout_1 (Dropout) (None, 32) 0 \n _________________________________________________________________\n dense_2 (Dense) (None, 1) 33 \n =================================================================\n Total params: 10,516,597\n Trainable params: 338,497\n Non-trainable params: 10,178,100\n\nBut when i fit my model, \nmodel.fit(x_train, y_train, validation_data=(x_val, y_val), batch_size = batch_size, epochs=num_epochs, verbose = True);\n\nIt is showing this error,\n\nTrain on 1044898 samples, validate on 261224 samples\nEpoch 1/5\n---------------------------------------------------------------------------\nInvalidArgumentError Traceback (most recent call last)\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _do_call(self, fn, *args)\n 1333 try:\n-> 1334 return fn(*args)\n 1335 except errors.OpError as e:\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _run_fn(feed_dict, fetch_list, target_list, options, run_metadata)\n 1316 # Ensure any changes to the graph are reflected in the runtime.\n-> 1317 self._extend_graph()\n 1318 return self._call_tf_sessionrun(\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _extend_graph(self)\n 1351 with self._graph._session_run_lock(): # pylint: disable=protected-access\n-> 1352 tf_session.ExtendSession(self._session)\n 1353 \n\nInvalidArgumentError: No OpKernel was registered to support Op 'CudnnRNN' with these attrs. Registered devices: [CPU,XLA_CPU], Registered kernels:\n ", "votes": 1 }, { "id": 457223, "postDate": "2019-01-17T05:17:10.920Z", "content": "you must enable the GPU is on. in the settings.
", "rawMarkdown": "you must enable the GPU is on. in the settings.", "replies": [ { "id": 457281, "postDate": "2019-01-17T07:25:53.017Z", "content": "I have enabled my GPU and then only i tried. But i got this weird error.
", "rawMarkdown": "I have enabled my GPU and then only i tried. But i got this weird error. " }, { "id": 457327, "postDate": "2019-01-17T08:32:58.483Z", "content": "did you use 'tf.device(\"/cpu:0\"):'?
", "rawMarkdown": "did you use 'tf.device(\"/cpu:0\"):'?" }, { "id": 458463, "postDate": "2019-01-19T17:44:41.217Z", "content": "What is that ? And I opened a new Kernel and the gpu works there. Thanks.
", "rawMarkdown": "What is that ? And I opened a new Kernel and the gpu works there. Thanks." } ] }, { "id": 456698, "postDate": "2019-01-16T10:26:50.900Z", "content": "Did you enable GPU kernel?
", "rawMarkdown": "Did you enable GPU kernel?", "replies": [ { "id": 457193, "postDate": "2019-01-17T04:07:52.003Z", "content": "Yeah.
", "rawMarkdown": "Yeah. " } ] }, { "id": 456707, "postDate": "2019-01-16T10:44:37.223Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 457194, "postDate": "2019-01-17T04:08:24.877Z", "content": "Can you suggest me on how to make this error disappear ?
", "rawMarkdown": "Can you suggest me on how to make this error disappear ?\n" } ] } ], "comments": [ { "id": 457223, "author_name": "Bai", "author_url": "", "post_date": "2019-01-17T05:17:10.920000", "content": "you must enable the GPU is on. in the settings.
", "votes": 0, "replies": [ { "id": 457281, "author_name": "Ajay Hayagreeve", "author_url": "", "post_date": "2019-01-17T07:25:53.017000", "content": "I have enabled my GPU and then only i tried. But i got this weird error.
", "votes": 0, "replies": [] }, { "id": 457327, "author_name": "Bai", "author_url": "", "post_date": "2019-01-17T08:32:58.483000", "content": "did you use 'tf.device(\"/cpu:0\"):'?
", "votes": 0, "replies": [] }, { "id": 458463, "author_name": "Ajay Hayagreeve", "author_url": "", "post_date": "2019-01-19T17:44:41.217000", "content": "What is that ? And I opened a new Kernel and the gpu works there. Thanks.
", "votes": 0, "replies": [] } ] }, { "id": 456698, "author_name": "Psi", "author_url": "", "post_date": "2019-01-16T10:26:50.900000", "content": "Did you enable GPU kernel?
", "votes": 0, "replies": [ { "id": 457193, "author_name": "Ajay Hayagreeve", "author_url": "", "post_date": "2019-01-17T04:07:52.003000", "content": "Yeah.
", "votes": 0, "replies": [] } ] }, { "id": 456707, "author_name": "", "author_url": "", "post_date": "2019-01-16T10:44:37.223000", "content": "", "votes": 0, "replies": [ { "id": 457194, "author_name": "Ajay Hayagreeve", "author_url": "", "post_date": "2019-01-17T04:08:24.877000", "content": "Can you suggest me on how to make this error disappear ?
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "456661": "Hi, \n In settings I have enabled my GPU. \n I was trying to implement the CuDNNGRU keras on kaggle. This is my model,\n\n inp = kl.Input(shape=(max_words,))\n m = kl.Embedding(input_dim = cl+1, output_dim = EMBEDDING_DIM, weights=[embedding_matrix], trainable = False)(inp)\n #del(embedding_matrix);\n m = kl.SpatialDropout1D(rate = 0.25)(m)\n m = kl.Bidirectional(kl.CuDNNGRU(128, kernel_initializer=glorot_normal(seed=1029), \n recurrent_initializer=orthogonal(gain=1.0, seed=1029), name = 'bi_gru'))(m)\n m = kl.Dense(32, activation=\"relu\")(m)\n m = kl.Dropout(0.2)(m)\n m = kl.Dense(1, activation=\"sigmoid\")(m)\n model = Model(inputs=inp, outputs=m)\n \n model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n print(model.summary())\n \n num_epochs = 5; batchSize = 128;\n\nAfter this this is my output.\n\n (33927, 300)\n _________________________________________________________________\n Layer (type) Output Shape Param # \n =================================================================\n input_1 (InputLayer) (None, 52) 0 \n _________________________________________________________________\n embedding_1 (Embedding) (None, 52, 300) 10178100 \n _________________________________________________________________\n spatial_dropout1d_1 (Spatial (None, 52, 300) 0 \n _________________________________________________________________\n bidirectional_1 (Bidirection (None, 256) 330240 \n _________________________________________________________________\n dense_1 (Dense) (None, 32) 8224 \n _________________________________________________________________\n dropout_1 (Dropout) (None, 32) 0 \n _________________________________________________________________\n dense_2 (Dense) (None, 1) 33 \n =================================================================\n Total params: 10,516,597\n Trainable params: 338,497\n Non-trainable params: 10,178,100\n\nBut when i fit my model, \nmodel.fit(x_train, y_train, validation_data=(x_val, y_val), batch_size = batch_size, epochs=num_epochs, verbose = True);\n\nIt is showing this error,\n\nTrain on 1044898 samples, validate on 261224 samples\nEpoch 1/5\n---------------------------------------------------------------------------\nInvalidArgumentError Traceback (most recent call last)\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _do_call(self, fn, *args)\n 1333 try:\n-> 1334 return fn(*args)\n 1335 except errors.OpError as e:\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _run_fn(feed_dict, fetch_list, target_list, options, run_metadata)\n 1316 # Ensure any changes to the graph are reflected in the runtime.\n-> 1317 self._extend_graph()\n 1318 return self._call_tf_sessionrun(\n\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py in _extend_graph(self)\n 1351 with self._graph._session_run_lock(): # pylint: disable=protected-access\n-> 1352 tf_session.ExtendSession(self._session)\n 1353 \n\nInvalidArgumentError: No OpKernel was registered to support Op 'CudnnRNN' with these attrs. Registered devices: [CPU,XLA_CPU], Registered kernels:\n ", "457223": "you must enable the GPU is on. in the settings.", "456698": "Did you enable GPU kernel?", "456707": "" } }