The program keeps reporting that it encounters embedding size not of 1024

[programmingnovicetao]

i used laser_encoders launched by facebook to embed my files, and i also checked the embeded files before running vecalign and confirmed that my files are made up of embedding size of 1024, but once i run vecalign, the program keeps reporting wrong embedding size error, i do not know what went wrong. please help me.

[programmingnovicetao]

the program i wrote would remind me the embedding size of my overlap files after finishing embedding, and you can see that my embeded files are all made up of embedding size of 1024, but vecalign reported the wrong size error again and again.

[thompsonb]

You either need to change this line to be compatible with your format:

vecalign/dp_utils.py

Line 63 in ca96a30

line_embeddings = np.fromfile(embed_file, dtype=np.float32, count=-1)

Or write your embeddings out in the format it is expecting. I think if you change your np.savetxt to ndarray.tofile (and maybe convert to np.float32 first) it should work.