What is model training vs inference?

Train vs infer

Training adjusts model weights on dataset (GPU-heavy, offline). Inference runs forward pass on new input in production (latency-sensitive).

offline: data → train → model artifact → online: request → infer → response