我训练了一个由ModelCheckpoint回调保存的模型。 我加载了它，并使用keras.Model.predict运行了预测，但是出现了“与矩阵大小不兼容”的错误，如下所示。
我检查了要进行预测的数据的形状正确，并且确实如此。
有什么建议吗？
代码
print("***dataset:")
print(dataset)
print("***Show shape")
iterator = dataset.make_one_shot_iterator()
next_batch = iterator.get_next()
try:
while True:
data = session.run(next_batch)
print(data.shape)
except tf.errors.OutOfRangeError:
pass
print("***Load model and predict")
model = tf.keras.models.load_model(model_file)
model.summary()
predictions = model.predict(dataset) # Matrix size-incompatible error
输出
***dataset:
<DatasetV1Adapter shapes: (?, 1, 64, ?), types: tf.float32>
***Show shape
(1, 1, 64, 169)
(1, 1, 64, 169)
(1, 1, 64, 169)
...
(1, 1, 64, 169)
(1, 1, 64, 169)
(1, 1, 64, 169)
(1, 1, 64, 169)
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
mels (InputLayer) [(32, 1, 64, 169)] 0
_________________________________________________________________
l1_conv (Conv2D) (32, 32, 62, 167) 288
_________________________________________________________________
l1_bn (BatchNormalization) (32, 32, 62, 167) 96
_________________________________________________________________
l1 (Activation) (32, 32, 62, 167) 0
_________________________________________________________________
l1_mp (MaxPooling2D) (32, 32, 30, 83) 0
_________________________________________________________________
l2_conv (Conv2D) (32, 32, 28, 81) 9216
_________________________________________________________________
l2_bn (BatchNormalization) (32, 32, 28, 81) 96
_________________________________________________________________
l2 (Activation) (32, 32, 28, 81) 0
_________________________________________________________________
l2_mp (MaxPooling2D) (32, 32, 13, 40) 0
_________________________________________________________________
l3_conv (Conv2D) (32, 32, 11, 38) 9216
_________________________________________________________________
l3_bn (BatchNormalization) (32, 32, 11, 38) 96
_________________________________________________________________
l3 (Activation) (32, 32, 11, 38) 0
_________________________________________________________________
l3_mp (MaxPooling2D) (32, 32, 5, 18) 0
_________________________________________________________________
flatten (Flatten) (32, 2880) 0
_________________________________________________________________
logits (Dense) (32, 100) 288100
_________________________________________________________________
dense (Dense) (32, 10) 1010
=================================================================
Total params: 308,118
Trainable params: 307,926
Non-trainable params: 192
***Load model and predict
InvalidArgumentError Traceback (most recent call last)
<ipython-input-14-edec6ee91517> in <module>
----> 1 predict('/home/jul/data/xenocanto/audio/wav_22050hz_MLR/XC164420.M.wav', '/home/jul/data/ingerop/subset_1572008350/features/actdet_config.json', '/home/jul/data/ingerop/subset_1572008350/features/featex_config.json', '/home/jul/data/ingerop/subset_1572008350/run_1572428779/models/model.05-0.92.h5')
~/dev/phaunos_ml/phaunos_ml/experiments/ingerop_prediction.py in predict(audio_filename, actdet_cfg_file, featex_cfg_file, model_file)
106 model.summary()
107
--> 108 predictions = model.predict(dataset)
109
110 return predictions
~/.miniconda3/envs/phaunos_ml/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py in predict(self, x, batch_size, verbose, steps, callbacks, max_queue_size, workers, use_multiprocessing)
1076 verbose=verbose,
1077 steps=steps,
-> 1078 callbacks=callbacks)
1079
1080 def reset_metrics(self):
~/.miniconda3/envs/phaunos_ml/lib/python3.6/site-packages/tensorflow/python/keras/engine/training_arrays.py in model_iteration(model, inputs, targets, sample_weights, batch_size, epochs, verbose, callbacks, val_inputs, val_targets, val_sample_weights, shuffle, initial_epoch, steps_per_epoch, validation_steps, validation_freq, mode, validation_in_fit, prepared_feed_values_from_dataset, steps_name, **kwargs)
272 # `ins` can be callable in tf.distribute.Strategy + eager case.
273 actual_inputs = ins() if callable(ins) else ins
--> 274 batch_outs = f(actual_inputs)
275 except errors.OutOfRangeError:
276 if is_dataset:
~/.miniconda3/envs/phaunos_ml/lib/python3.6/site-packages/tensorflow/python/keras/backend.py in __call__(self, inputs)
3290
3291 fetched = self._callable_fn(*array_vals,
-> 3292 run_metadata=self.run_metadata)
3293 self._call_fetch_callbacks(fetched[-len(self._fetches):])
3294 output_structure = nest.pack_sequence_as(
~/.miniconda3/envs/phaunos_ml/lib/python3.6/site-packages/tensorflow/python/client/session.py in __call__(self, *args, **kwargs)
1456 ret = tf_session.TF_SessionRunCallable(self._session._session,
1457 self._handle, args,
-> 1458 run_metadata_ptr)
1459 if run_metadata:
1460 proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)
InvalidArgumentError: 2 root error(s) found.
(0) Invalid argument: Matrix size-incompatible: In[0]: [32,90], In[1]: [2880,100]
[[{{node logits_9/MatMul}}]]
(1) Invalid argument: Matrix size-incompatible: In[0]: [32,90], In[1]: [2880,100]
[[{{node logits_9/MatMul}}]]
[[dense_9/Sigmoid/_2867]]
0 successful operations.
0 derived errors ignored.