Remove quantization_config from the APIs for now

llama_toolchain/agentic_system/api/datatypes.py

@@ -151,8 +151,6 @@ class AgenticSystemInstanceConfig(BaseModel):
     input_shields: Optional[List[ShieldDefinition]] = Field(default_factory=list)
     output_shields: Optional[List[ShieldDefinition]] = Field(default_factory=list)
 
-    quantization_config: Optional[QuantizationConfig] = None
-
     # if you completely want to replace the messages prefixed by the system,
     # this is debug only
     debug_prefix_messages: Optional[List[Message]] = Field(default_factory=list)

llama_toolchain/agentic_system/client.py

@@ -135,7 +135,6 @@ async def run_main(host: str, port: int):
             available_tools=tool_definitions,
             input_shields=[],
             output_shields=[],
-            quantization_config=None,
             debug_prefix_messages=[],
             tool_prompt_format=ToolPromptFormat.json,
         ),

llama_toolchain/inference/api/endpoints.py

@@ -19,7 +19,6 @@ class CompletionRequest(BaseModel):
 
     stream: Optional[bool] = False
     logprobs: Optional[LogProbConfig] = None
-    quantization_config: Optional[QuantizationConfig] = None
 
 
 @json_schema_type
@@ -43,7 +42,6 @@ class BatchCompletionRequest(BaseModel):
     content_batch: List[InterleavedTextAttachment]
     sampling_params: Optional[SamplingParams] = SamplingParams()
     logprobs: Optional[LogProbConfig] = None
-    quantization_config: Optional[QuantizationConfig] = None
 
 
 @json_schema_type
@@ -62,7 +60,6 @@ class ChatCompletionRequest(BaseModel):
 
     stream: Optional[bool] = False
     logprobs: Optional[LogProbConfig] = None
-    quantization_config: Optional[QuantizationConfig] = None
 
 
 @json_schema_type
@@ -88,7 +85,6 @@ class BatchChatCompletionRequest(BaseModel):
     available_tools: Optional[List[ToolDefinition]] = Field(default_factory=list)
 
     logprobs: Optional[LogProbConfig] = None
-    quantization_config: Optional[QuantizationConfig] = None
 
 
 @json_schema_type