Privacy Filter Timing
openai/privacy-filter · bf16 · 2 CPU threads · torch 2.14.0+cpu · model loaded (1399M params, RSS 3512 MB)
Model load (one-off, at process start 2026-10-04 07:28:54)
| tokenizer | 2551 ms |
| weights (safetensors → bf16) | 97 ms |
| read all weights into memory and pin them (2669 MB locked) | 246 ms |
| Viterbi transition matrices | 4.2 ms |
| warm-up forward passes | 3107 ms |
| RSS added by model | 3039 MB |