Bugfix: missing response_format for functionary and llava chat handlers

Author: abetlen

llama_cpp/llama_chat_format.py

@@ -318,10 +318,11 @@ def basic_create_chat_completion(
                 stop = [] if stop is None else [stop] if isinstance(stop, str) else stop
                 rstop = result.stop if isinstance(result.stop, list) else [result.stop]
                 stop = stop + rstop
-            
+
             if response_format is not None and response_format["type"] == "json_object":
-                print("hello world")
-                grammar = llama_grammar.LlamaGrammar.from_string(llama_grammar.JSON_GBNF)
+                grammar = llama_grammar.LlamaGrammar.from_string(
+                    llama_grammar.JSON_GBNF
+                )
 
             completion_or_chunks = llama.create_completion(
                 prompt=prompt,
@@ -577,6 +578,7 @@ def functionary_chat_handler(
     top_k: int = 40,
     stream: bool = False,
     stop: Optional[Union[str, List[str]]] = [],
+    response_format: Optional[llama_types.ChatCompletionRequestResponseFormat] = None,
     max_tokens: int = 256,
     presence_penalty: float = 0.0,
     frequency_penalty: float = 0.0,
@@ -753,6 +755,10 @@ def message_to_str(msg: llama_types.ChatCompletionRequestMessage):
     assert isinstance(function_call, str)
     assert stream is False  # TODO: support stream mode
 
+    if response_format is not None and response_format["type"] == "json_object":
+        with suppress_stdout_stderr(disable=llama.verbose):
+            grammar = llama_grammar.LlamaGrammar.from_string(llama_grammar.JSON_GBNF)
+
     return llama_types.CreateChatCompletionResponse(
         id="chat" + completion["id"],
         object="chat.completion",
@@ -785,11 +791,11 @@ def __init__(self, clip_model_path: str, verbose: bool = False):
         self._llava_cpp = llava_cpp
         self.clip_model_path = clip_model_path
         self.verbose = verbose
-        self._clip_free = self._llava_cpp._libllava.clip_free # type: ignore
+        self._clip_free = self._llava_cpp._libllava.clip_free  # type: ignore
 
         with suppress_stdout_stderr(disable=self.verbose):
             self.clip_ctx = self._llava_cpp.clip_model_load(
-                self.clip_model_path.encode(), 0 
+                self.clip_model_path.encode(), 0
             )
 
     def __del__(self):
@@ -825,6 +831,9 @@ def __call__(
         top_k: int = 40,
         stream: bool = False,
         stop: Optional[Union[str, List[str]]] = [],
+        response_format: Optional[
+            llama_types.ChatCompletionRequestResponseFormat
+        ] = None,
         max_tokens: int = 256,
         presence_penalty: float = 0.0,
         frequency_penalty: float = 0.0,
@@ -851,7 +860,6 @@ def __call__(
             if system_prompt != ""
             else "A chat between a curious human and an artificial intelligence assistant.  The assistant gives helpful, detailed, and polite answers to the human's questions."
         )
-        system_prompt = "A chat between a curious human and an artificial intelligence assistant.  The assistant gives helpful, detailed, and polite answers to the human's questions."
         user_role = "\nUSER:"
         assistant_role = "\nASSISTANT:"
         llama.reset()
@@ -890,11 +898,13 @@ def __call__(
                                 ctypes.c_ubyte * len(data_array)
                             ).from_buffer(data_array)
                             with suppress_stdout_stderr(disable=self.verbose):
-                                embed = self._llava_cpp.llava_image_embed_make_with_bytes(
-                                    ctx_clip=self.clip_ctx,
-                                    n_threads=llama.context_params.n_threads,
-                                    image_bytes=c_ubyte_ptr,
-                                    image_bytes_length=len(image_bytes),
+                                embed = (
+                                    self._llava_cpp.llava_image_embed_make_with_bytes(
+                                        ctx_clip=self.clip_ctx,
+                                        n_threads=llama.context_params.n_threads,
+                                        image_bytes=c_ubyte_ptr,
+                                        image_bytes_length=len(image_bytes),
+                                    )
                                 )
                             try:
                                 n_past = ctypes.c_int(llama.n_tokens)
@@ -917,9 +927,17 @@ def __call__(
                         f"ASSISTANT: {message['content']}".encode("utf8"), add_bos=False
                     )
                 )
+                assert llama.n_ctx() >= llama.n_tokens
         llama.eval(llama.tokenize(f"{assistant_role}".encode("utf8"), add_bos=False))
+        assert llama.n_ctx() >= llama.n_tokens
 
-        prompt = llama.input_ids[:llama.n_tokens].tolist()
+        prompt = llama.input_ids[: llama.n_tokens].tolist()
+
+        if response_format is not None and response_format["type"] == "json_object":
+            with suppress_stdout_stderr(disable=self.verbose):
+                grammar = llama_grammar.LlamaGrammar.from_string(
+                    llama_grammar.JSON_GBNF
+                )
 
         return _convert_completion_to_chat(
             llama.create_completion(