ValueError：未知度量函数：余弦

Question

i have been getting valueError issue. Currently using python3.9.11., keras2.8.

        if loss_init=="r2":
        parallel_model.compile(loss=custom_r2_loss, optimizer=opt,  metrics=['mse','mae', 'mape', 'cosine','acc', custom_r2_loss])
    elif loss_init =="wmae":
        parallel_model.compile(loss=custom_wmae_loss, optimizer=opt,  metrics=['mse','mae', 'mape', 'cosine','acc', custom_wmae_loss])
    else:
        parallel_model.compile(loss=loss_init, optimizer=opt,  metrics=['mse','mae', 'mape', 'cosine','acc']) 
    

and it gives this error: ValueError: Unknown metric function: cosine. Please ensure this object is passed to the custom_objects argument. See https://www.tensorflow.org/guide/keras/save_and_serialize#registering_the_custom_object for details.

Here is the full traceback: ValueError: in user code:

File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/training.py", line 1021, in train_function  *
    return step_function(self, iterator)
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/training.py", line 1010, in step_function  **
    outputs = model.distribute_strategy.run(run_step, args=(data,))
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/training.py", line 1000, in run_step  **
    outputs = model.train_step(data)
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/training.py", line 864, in train_step
    return self.compute_metrics(x, y, y_pred, sample_weight)
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/training.py", line 957, in compute_metrics
    self.compiled_metrics.update_state(y, y_pred, sample_weight)
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/compile_utils.py", line 438, in update_state
    self.build(y_pred, y_true)
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/compile_utils.py", line 358, in build
    self._metrics = tf.__internal__.nest.map_structure_up_to(y_pred, self._get_metric_objects,
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/compile_utils.py", line 484, in _get_metric_objects
    return [self._get_metric_object(m, y_t, y_p) for m in metrics]
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/compile_utils.py", line 484, in <listcomp>
    return [self._get_metric_object(m, y_t, y_p) for m in metrics]
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/engine/compile_utils.py", line 503, in _get_metric_object
    metric_obj = metrics_mod.get(metric)
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/metrics.py", line 4262, in get
    return deserialize(str(identifier))
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/metrics.py", line 4218, in deserialize
    return deserialize_keras_object(
File "/Users/neslihanyuksel/opt/anaconda3/envs/hipokrat/lib/python3.9/site-packages/keras/utils/generic_utils.py", line 709, in deserialize_keras_object
    raise ValueError(

ValueError: Unknown metric function: cosine. Please ensure this object is passed to the `custom_objects` argument. See https://www.tensorflow.org/guide/keras/save_and_serialize#registering_the_custom_object for details.




ValueError: Unknown metric function: cosine. Please ensure this object is passed to the `custom_objects` argument. See https://www.tensorflow.org/guide/keras/save_and_serialize#registering_the_custom_object for details.

full code is https://github.com/lrsoenksen/CL_RNA_SynthBio/blob/master/1_DL_CNN_2D_toehold_V1.ipynb

Answer 1

Replacing cosine with cosine_similarity will help. keras.metrics.CosineSimilarity computes the cosine similarity between the labels and predictions.

parallel_model.compile(loss=custom_r2_loss, optimizer=opt,  metrics=['mse','mae', 'mape', 'cosine_similarity','acc', custom_r2_loss])

You can find the reference here . Thank you!