Wrong model outputs #4
Description
Hi @JappyPing @tianlt ,
I have been following the demo from the previous version of this repository, but I encountered some issues with the model outputs. Specifically, I obtained the following mutated sequence (showing a partial output):
LTIEEQAKTFLDKFNHEAEDLFYQSSLASWNYNTNITEENVQNMNNAGDKLLLLKKKKKKKEEMYPLQEIQNLTVKLQLQALQQNGSSEEEEEEEEEKNTILNTMSTIYSTGKVCNPDNPQECLLLEPGLNEIMANSLDKKKKKKKKKKKKKKKKKKKKKKYEEYVVLKNEMA...
As you can see, there are repetitive mutations like KKKKK and LLLL, which seem quite abnormal. More importantly, these mutations are not limited to the RBD region, which contradicts expectations.
For reference, here is the original sequence before mutation:
STIEEQAKTFLDKFNHEAEDLFYQSSLASWNYNTNITEENVQNMNNAGDKWSAFLKEQSTLAQMYPLQEIQNLTVKLQLQALQQNGSSVLSEDKSKRLNTILNTMSTIYSTGKVCNPDNPQECLLLEPGLNEIMANSLDYNERLWAWESWRSEVGKQLRPLYEEYVVLKNEMA...
Additionally, this is the rbd_mask I used:
00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000...
As the rbd_mask is all zeros for this part of the sequence, no mutations should have occurred in these regions. However, unexpected mutations still appeared.
To provide more context, here are the model inputs I used, all taken from the previous version of the repository as the current version does NOT include demo scripts or data:
file_names = ['pre.pkl', 'rbd.pkl', 'same.pkl', 'x.pkl', 'y.pkl', 'z.pkl', 'noise.pkl']
input_data = [pre, rbd, same, x, y, z, noise]
pre: tf.Tensor: shape=(1, 3924, 20), dtype=float32
rbd: tf.Tensor: shape=(1, 3924, 1)
same: tf.Tensor: shape=(1, 3924, 1)
x: tf.Tensor: shape=(1, 3924, 1)
y: tf.Tensor: shape=(1, 3924, 1)
z: tf.Tensor: shape=(1, 3924, 1)
noise: tf.Tensor: shape=(1, 3924, 1)
The model parameters were loaded from 'deepdirect_framework/model_i_weights.h5'
Could you kindly help with this issue? It would be really helpful if you could either update the code or provide clearer instructions on how to properly run the deepdirect model.
Looking forward to your response, and thanks in advance for your help!