support for arbitrary kwargs for llamacpp (#8727)

llamacpp params (per their own code) are unstable, so instead of
adding/deleting them constantly adding a model_kwargs parameter that
allows for arbitrary additional kwargs

cc @jsjolund and @zacps re #8599 and #8704

libs/langchain/langchain/llms/llamacpp.py

@@ -6,6 +6,8 @@ from pydantic import Field, root_validator
 from langchain.callbacks.manager import CallbackManagerForLLMRun
 from langchain.llms.base import LLM
 from langchain.schema.output import GenerationChunk
+from langchain.utils import get_pydantic_field_names
+from langchain.utils.utils import build_extra_kwargs
 
 logger = logging.getLogger(__name__)
 
@@ -106,6 +108,9 @@ class LlamaCpp(LLM):
     rope_freq_base: float = 10000.0
     """Base frequency for rope sampling."""
 
+    model_kwargs: Dict[str, Any] = Field(default_factory=dict)
+    """Any additional parameters to pass to llama_cpp.Llama."""
+
     streaming: bool = True
     """Whether to stream the results, token by token."""
 
@@ -139,6 +144,8 @@ class LlamaCpp(LLM):
         if values["n_gpu_layers"] is not None:
             model_params["n_gpu_layers"] = values["n_gpu_layers"]
 
+        model_params.update(values["model_kwargs"])
+
         try:
             from llama_cpp import Llama
 
@@ -157,6 +164,16 @@ class LlamaCpp(LLM):
 
         return values
 
+    @root_validator(pre=True)
+    def build_model_kwargs(cls, values: Dict[str, Any]) -> Dict[str, Any]:
+        """Build extra kwargs from additional params that were passed in."""
+        all_required_field_names = get_pydantic_field_names(cls)
+        extra = values.get("model_kwargs", {})
+        values["model_kwargs"] = build_extra_kwargs(
+            extra, values, all_required_field_names
+        )
+        return values
+
     @property
     def _default_params(self) -> Dict[str, Any]:
         """Get the default parameters for calling llama_cpp."""

libs/langchain/langchain/llms/openai.py

@@ -30,6 +30,7 @@ from langchain.llms.base import BaseLLM, create_base_retry_decorator
 from langchain.schema import Generation, LLMResult
 from langchain.schema.output import GenerationChunk
 from langchain.utils import get_from_dict_or_env, get_pydantic_field_names
+from langchain.utils.utils import build_extra_kwargs
 
 logger = logging.getLogger(__name__)
 
@@ -215,25 +216,9 @@ class BaseOpenAI(BaseLLM):
         """Build extra kwargs from additional params that were passed in."""
         all_required_field_names = get_pydantic_field_names(cls)
         extra = values.get("model_kwargs", {})
-        for field_name in list(values):
+        values["model_kwargs"] = build_extra_kwargs(
-            if field_name in extra:
+            extra, values, all_required_field_names
-                raise ValueError(f"Found {field_name} supplied twice.")
+        )
-            if field_name not in all_required_field_names:
-                warnings.warn(
-                    f"""WARNING! {field_name} is not default parameter.
-                    {field_name} was transferred to model_kwargs.
-                    Please confirm that {field_name} is what you intended."""
-                )
-                extra[field_name] = values.pop(field_name)
-
-        invalid_model_kwargs = all_required_field_names.intersection(extra.keys())
-        if invalid_model_kwargs:
-            raise ValueError(
-                f"Parameters {invalid_model_kwargs} should be specified explicitly. "
-                f"Instead they were passed in as part of `model_kwargs` parameter."
-            )
-
-        values["model_kwargs"] = extra
         return values
 
     @root_validator()

libs/langchain/langchain/utils/utils.py

@@ -2,8 +2,9 @@
 import contextlib
 import datetime
 import importlib
+import warnings
 from importlib.metadata import version
-from typing import Any, Callable, Optional, Set, Tuple
+from typing import Any, Callable, Dict, Optional, Set, Tuple
 
 from packaging.version import parse
 from requests import HTTPError, Response
@@ -122,7 +123,7 @@ def check_package_version(
         )
 
 
-def get_pydantic_field_names(pydantic_cls: Any) -> Set:
+def get_pydantic_field_names(pydantic_cls: Any) -> Set[str]:
     """Get field names, including aliases, for a pydantic class.
 
     Args:
@@ -133,3 +134,30 @@ def get_pydantic_field_names(pydantic_cls: Any) -> Set:
         if field.has_alias:
             all_required_field_names.add(field.alias)
     return all_required_field_names
+
+
+def build_extra_kwargs(
+    extra_kwargs: Dict[str, Any],
+    values: Dict[str, Any],
+    all_required_field_names: Set[str],
+) -> Dict[str, Any]:
+    """"""
+    for field_name in list(values):
+        if field_name in extra_kwargs:
+            raise ValueError(f"Found {field_name} supplied twice.")
+        if field_name not in all_required_field_names:
+            warnings.warn(
+                f"""WARNING! {field_name} is not default parameter.
+                {field_name} was transferred to model_kwargs.
+                Please confirm that {field_name} is what you intended."""
+            )
+            extra_kwargs[field_name] = values.pop(field_name)
+
+    invalid_model_kwargs = all_required_field_names.intersection(extra_kwargs.keys())
+    if invalid_model_kwargs:
+        raise ValueError(
+            f"Parameters {invalid_model_kwargs} should be specified explicitly. "
+            f"Instead they were passed in as part of `model_kwargs` parameter."
+        )
+
+    return extra_kwargs

libs/langchain/tests/integration_tests/llms/test_llamacpp.py

@@ -4,6 +4,8 @@ import os
 from typing import Generator
 from urllib.request import urlretrieve
 
+import pytest
+
 from langchain.llms import LlamaCpp
 
 from tests.unit_tests.callbacks.fake_callback_handler import FakeCallbackHandler
@@ -68,3 +70,19 @@ def test_llamacpp_streaming_callback() -> None:
     )
     llm("Q: Can you count to 10? A:'1, ")
     assert callback_handler.llm_streams <= MAX_TOKENS + OFF_BY_ONE
+
+
+def test_llamacpp_model_kwargs() -> None:
+    llm = LlamaCpp(model_path=get_model(), model_kwargs={"n_gqa": None})
+    assert llm.model_kwargs == {"n_gqa": None}
+
+
+def test_llamacpp_invalid_model_kwargs() -> None:
+    with pytest.raises(ValueError):
+        LlamaCpp(model_path=get_model(), model_kwargs={"n_ctx": 1024})
+
+
+def test_llamacpp_incorrect_field() -> None:
+    with pytest.warns(match="not default parameter"):
+        llm = LlamaCpp(model_path=get_model(), n_gqa=None)
+    llm.model_kwargs == {"n_gqa": None}

libs/langchain/tests/unit_tests/llms/test_openai.py

@@ -22,6 +22,12 @@ def test_openai_model_param() -> None:
     assert llm.model_name == "foo"
 
 
+@pytest.mark.requires("openai")
+def test_openai_model_kwargs() -> None:
+    llm = OpenAI(model_kwargs={"foo": "bar"})
+    assert llm.model_kwargs == {"foo": "bar"}
+
+
 @pytest.mark.requires("openai")
 def test_openai_invalid_model_kwargs() -> None:
     with pytest.raises(ValueError):