UltraMS Search

The spectrum-to-spectrum contrastive model selected for the UltraAtlas application (best.pt). encode(...).embedding returns its normalized 512-dimensional projection of the UltraMS encoder's CLS embedding. Use this representation to rank spectra in a supplied reference library by cosine similarity.

Output Value
Embedding dimension 512
Maximum spectral peaks 150
Python model name "search"

model.pt contains model weights and inference configuration and can be loaded with UltraMS 0.3.2.

python -m pip install ultrams
from ultrams import UltraMS

model = UltraMS.from_pretrained("search")
embedding = model.encode(
    mz=[100.1, 121.1, 150.0], intensity=[20, 100, 35], precursor_mz=301.2
).embedding
print(embedding.shape)  # (512,)

Supply measured spectral peak m/z and intensity arrays plus precursor-ion m/z. UltraMS requires at least three spectral peaks with positive m/z, normalizes intensities by their maximum when positive, and retains the 150 most intense spectral peaks when needed. See the input format, model selection, and spectrum search example.

Use return_peaks=True in model.encode(...) to obtain final encoder embeddings aligned with the retained spectral peaks. The three released checkpoints are in the UltraMS model family.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Collection including dsadd4/UltraMS-Search