Model precision

#1
by Jrd100 - opened

When I deploy this model in ollama, it seems to use half precision instead of full precision:

prithivMLmods/SmolLM2-135M-F32-GGUF

ollama list
NAME ID SIZE MODIFIED
hf.co/prithivMLmods/SmolLM2-135M-F32-GGUF:latest 5fce903808f1 270 MB About a minute ago

for a 135M parameter model, full precision should be 4x135 = 540MB and not 270MB.

Do you know why this discrepancy?
Thanks

Sign up or log in to comment