Model: Attach request ID to logs

If multiple logs come in at once, track which log corresponds to
which request.

Signed-off-by: kingbri <bdashore3@proton.me>

backends/exllamav2/model.py

@@ -1210,7 +1210,7 @@ class ExllamaV2Container:
 
                     # Second yield if eos is true
                     if result.get("eos"):
-                        log_response(full_response)
+                        log_response(request_id, full_response)
 
                         eos_reason = result.get("eos_reason")
                         finish_reason = (
@@ -1271,6 +1271,7 @@ class ExllamaV2Container:
             # Log the metrics if present
             if metrics_result:
                 log_metrics(
+                    request_id,
                     metrics_result.get("time_enqueued"),
                     metrics_result.get("prompt_tokens"),
                     metrics_result.get("cached_tokens"),

common/gen_logging.py

@@ -64,14 +64,18 @@ def log_prompt(prompt: str, request_id: str, negative_prompt: Optional[str]):
             logger.info(f"Negative Prompt: {formatted_negative_prompt}\n")
 
 
-def log_response(response: str):
+def log_response(request_id: str, response: str):
     """Logs the response to console."""
     if PREFERENCES.prompt:
         formatted_response = "\n" + response
-        logger.info(f"Response: {formatted_response if response else 'Empty'}\n")
+        logger.info(
+            f"Response (ID: {request_id}): "
+            f"{formatted_response if response else 'Empty'}\n"
+        )
 
 
 def log_metrics(
+    request_id: str,
     queue_time: float,
     prompt_tokens: int,
     cached_tokens: int,
@@ -82,7 +86,7 @@ def log_metrics(
     max_seq_len: int,
 ):
     initial_response = (
-        f"Metrics: {generated_tokens} tokens generated in "
+        f"Metrics (ID: {request_id}): {generated_tokens} tokens generated in "
         f"{round(queue_time + prompt_time + generate_time, 2)} seconds"
     )
     itemization = []