Config: replace disable_output_chunking flag with output_chunking

backends/exllamav3/model.py

@@ -250,8 +250,8 @@ class ExllamaV3Container(BaseModelContainer):
         self.chunk_size = self.adjust_chunk_size(user_chunk_size)
 
         # Output chunking
-        disable_output_chunking = unwrap(kwargs.get("disable_output_chunking"), False)
-        self.max_rq_tokens = None if disable_output_chunking else self.chunk_size
+        output_chunking = unwrap(kwargs.get("output_chunking"), True)
+        self.max_rq_tokens = self.chunk_size if output_chunking else None
 
         # Template setup
         self.prompt_template = await find_prompt_template(

common/config_models.py

@@ -265,12 +265,13 @@ class ModelConfig(BaseConfigModel):
         ),
         gt=0,
     )
-    disable_output_chunking: Optional[bool] = Field(
-        False,
+    output_chunking: Optional[bool] = Field(
+        True,
         description=(
-            "Disable output chunking (default: false).\n"
+            "Use output chunking (default: True)\n"
+            "Instead of allocating cache space for the entire completion at once, "
+            "allocate in chunks as needed.\n"
             "Used by EXL3 models only.\n"
-            "True, allocate space in the cache for the entire response with each request..\n"
         ),
     )
     max_batch_size: Optional[int] = Field(

config_sample.yml

@@ -133,10 +133,10 @@ model:
   # An ideal value is between 512 and 4096.
   chunk_size: 2048
 
-  # Disable output chunking (default: false)
+  # Use output chunking (default: True)
+  # Instead of allocating cache space for the entire completion at once, allocate in chunks as needed.
   # Used by EXL3 models only.
-  # If True, allocate space in the cache for the entire response with each request.
-  disable_output_chunking: false
+  output_chunking: true
 
   # Set the maximum number of prompts to process at one time (default: None/Automatic).
   # Automatically calculated if left blank.

endpoints/core/types/model.py

@@ -109,7 +109,7 @@ class ModelLoadRequest(BaseModel):
     )
     cache_mode: Optional[str] = None
     chunk_size: Optional[int] = None
-    disable_output_chunking: Optional[bool] = False
+    output_chunking: Optional[bool] = True
     prompt_template: Optional[str] = None
     vision: Optional[bool] = None