Logging: Clarify new vs cached tokens in prompt processing

2026-04-20 14:28:54 +00:00 · 2024-05-26 18:21:17 -07:00
parent 3dcae8b023
commit ce5e2ec8de
2 changed files with 8 additions and 2 deletions
--- a/backends/exllamav2/model.py
+++ b/backends/exllamav2/model.py
@@ -1125,6 +1125,7 @@ class ExllamaV2Container:
                        log_metrics(
                            result.get("time_enqueued"),
                            result.get("prompt_tokens"),
+                            result.get("cached_tokens"),
                            result.get("time_prefill"),
                            result.get("new_tokens"),
                            result.get("time_generate"),