EXPERIMENT 001

Sentiment
classification

Apparatus
hostinger-kerry · 2 vCPU
Model
distilbert-base-uncased
-finetuned-sst-2-english
Method
Three inference stacks, one
specimen, measured concurrently

01 Specimen

The same specimen is dispatched to all three apparatus at once. Only the inference stack differs.

02 Observation

A python

FastAPI
+ Python

Inference
PyTorch
Weights
safetensors

awaiting specimen

B rust → python

Rust / Axum
+ Python

Inference
PyTorch, unchanged
Weights
safetensors

awaiting specimen

C rust

Rust / Axum
+ rust-bert

Inference
tch / LibTorch
Weights
.ot native

awaiting specimen

03 Record

Control One sentence, one model, one machine — held constant across A, B and C.
Variable Inference stack only: Python · Rust fronting Python · Rust native.
Observation —
Conclusion —