I am trying to use the python library provided for computing TMK, and realized the vector value is a lot different than the TMK L1 features computed in C++. Is this expected?

For more details:
hasher = hashers.TMKL1(frame_hasher=PDQHashF())
hash1 = hasher.compute(filepath='../chair-22-sd-grey-bar.mp4', hash_format="vector")
The videos are from here: https://github.com/facebook/ThreatExchange/tree/main/tmk/sample-videos

While compare with the normalized feature generated from ThreatExchange TMK, the cosine distance between the two vectors is 1.17 (cosine similarity = -0.17).

I am trying to understand if this is expected, it's due to not using the correct parameters. Thank you!