feat: Add stopping_criteria to ChatFormatter, allow stopping on arbitrary token ids, fixes llama3 instruct

abetlen · abetlen · commit cc81afebf04d · 2024-04-20T00:00:53.000-04:00
diff --git a/llama_cpp/llama.py b/llama_cpp/llama.py
@@ -426,7 +426,10 @@ def __init__(
                     print(f"Using chat bos_token: {bos_token}", file=sys.stderr)
 
                 self.chat_handler = llama_chat_format.Jinja2ChatFormatter(
-                    template=template, eos_token=eos_token, bos_token=bos_token
+                    template=template,
+                    eos_token=eos_token,
+                    bos_token=bos_token,
+                    stop_token_ids=[eos_token_id],
                 ).to_chat_handler()
 
         if self.chat_format is None and self.chat_handler is None:
diff --git a/llama_cpp/llama_chat_format.py b/llama_cpp/llama_chat_format.py
@@ -10,6 +10,9 @@
 
 import jinja2
 
+import numpy as np
+import numpy.typing as npt
+
 import llama_cpp.llama as llama
 import llama_cpp.llama_types as llama_types
 import llama_cpp.llama_grammar as llama_grammar
@@ -150,6 +153,7 @@ class ChatFormatterResponse:
 
     prompt: str
     stop: Optional[Union[str, List[str]]] = None
+    stopping_criteria: Optional[llama.StoppingCriteriaList] = None
 
 
 class ChatFormatter(Protocol):
@@ -173,12 +177,14 @@ def __init__(
         eos_token: str,
         bos_token: str,
         add_generation_prompt: bool = True,
+        stop_token_ids: Optional[List[int]] = None,
     ):
         """A chat formatter that uses jinja2 templates to format the prompt."""
         self.template = template
         self.eos_token = eos_token
         self.bos_token = bos_token
         self.add_generation_prompt = add_generation_prompt
+        self.stop_token_ids = set(stop_token_ids) if stop_token_ids is not None else None
 
         self._environment = jinja2.Environment(
             loader=jinja2.BaseLoader(),
@@ -211,7 +217,16 @@ def raise_exception(message: str):
             tool_choice=tool_choice,
         )
 
-        return ChatFormatterResponse(prompt=prompt, stop=[self.eos_token])
+        stopping_criteria = None
+        if self.stop_token_ids is not None:
+            def stop_on_last_token(
+                    tokens: npt.NDArray[np.intc],
+                    logits: npt.NDArray[np.single]
+            ) -> bool:
+                return tokens[-1] in self.stop_token_ids
+            stopping_criteria = llama.StoppingCriteriaList([stop_on_last_token])
+
+        return ChatFormatterResponse(prompt=prompt, stop=[self.eos_token], stopping_criteria=stopping_criteria)
 
     def to_chat_handler(self) -> LlamaChatCompletionHandler:
         return chat_formatter_to_chat_completion_handler(self)
@@ -533,6 +548,10 @@ def chat_completion_handler(
             rstop = result.stop if isinstance(result.stop, list) else [result.stop]
             stop = stop + rstop
 
+        stopping_criteria = None
+        if result.stopping_criteria is not None:
+            stopping_criteria = result.stopping_criteria
+
         if response_format is not None and response_format["type"] == "json_object":
             grammar = _grammar_for_response_format(response_format, verbose=llama.verbose)
 
@@ -598,6 +617,7 @@ def chat_completion_handler(
             mirostat_eta=mirostat_eta,
             model=model,
             logits_processor=logits_processor,
+            stopping_criteria=stopping_criteria,
             grammar=grammar,
             logit_bias=logit_bias,
         )
