fix: adjust usage of vLLM for deprecation of 'disable_log_requests' (#4659)

Co-authored-by: Ryan McCormick <rmccormick@nvidia.com>

components/src/dynamo/vllm/args.py

@@ -464,7 +464,7 @@ def overwrite_args(config):
         # skip tokenizer initialisation.  Setting this to **False** avoids
         # a NoneType error when the processor accesses the tokenizer.
         "skip_tokenizer_init": False,
-        "disable_log_requests": True,
+        "enable_log_requests": False,
         "disable_log_stats": False,
     }
 

components/src/dynamo/vllm/main.py

@@ -296,7 +296,7 @@ def setup_vllm_engine(config, stat_logger=None):
         vllm_config=vllm_config,
         usage_context=usage_context,
         stat_loggers=factory,
-        disable_log_requests=engine_args.disable_log_requests,
+        enable_log_requests=engine_args.enable_log_requests,
         disable_log_stats=engine_args.disable_log_stats,
     )
 

examples/multimodal/components/worker.py

@@ -142,7 +142,7 @@ class VllmBaseWorker:
             vllm_config=vllm_config,
             usage_context=usage_context,
             stat_loggers=[self.stats_logger],
-            disable_log_requests=self.engine_args.disable_log_requests,
+            enable_log_requests=self.engine_args.enable_log_requests,
             disable_log_stats=self.engine_args.disable_log_stats,
         )
 

examples/multimodal/utils/args.py

@@ -155,7 +155,7 @@ def overwrite_args(config):
     defaults = {
         "task": "generate",
         "skip_tokenizer_init": False,
-        "disable_log_requests": True,
+        "enable_log_requests": False,
         "enable_prefix_caching": True,
         # KV routing relies on logging KV metrics
         "disable_log_stats": False,