Update LlamaCpp parameters (#2411)

Add `n_batch` and `last_n_tokens_size` parameters to the LlamaCpp class.
These parameters (epecially `n_batch`) significantly effect performance.
There's also a `verbose` flag that prints system timings on the `Llama`
class but I wasn't sure where to add this as it conflicts with (should
be pulled from?) the LLM base class.

langchain/embeddings/llamacpp.py

@@ -49,6 +49,10 @@ class LlamaCppEmbeddings(BaseModel, Embeddings):
     """Number of threads to use. If None, the number 
     of threads is automatically determined."""
 
+    n_batch: Optional[int] = Field(8, alias="n_batch")
+    """Number of tokens to process in parallel.
+    Should be a number between 1 and n_ctx."""
+
     class Config:
         """Configuration for this pydantic object."""
 
@@ -66,6 +70,7 @@ class LlamaCppEmbeddings(BaseModel, Embeddings):
         vocab_only = values["vocab_only"]
         use_mlock = values["use_mlock"]
         n_threads = values["n_threads"]
+        n_batch = values["n_batch"]
 
         try:
             from llama_cpp import Llama
@@ -80,6 +85,7 @@ class LlamaCppEmbeddings(BaseModel, Embeddings):
                 vocab_only=vocab_only,
                 use_mlock=use_mlock,
                 n_threads=n_threads,
+                n_batch=n_batch,
                 embedding=True,
             )
         except ImportError:

langchain/llms/llamacpp.py

@@ -53,6 +53,10 @@ class LlamaCpp(LLM, BaseModel):
     """Number of threads to use. 
     If None, the number of threads is automatically determined."""
 
+    n_batch: Optional[int] = Field(8, alias="n_batch")
+    """Number of tokens to process in parallel.
+    Should be a number between 1 and n_ctx."""
+
     suffix: Optional[str] = Field(None)
     """A suffix to append to the generated text. If None, no suffix is appended."""
 
@@ -80,6 +84,9 @@ class LlamaCpp(LLM, BaseModel):
     top_k: Optional[int] = 40
     """The top-k value to use for sampling."""
 
+    last_n_tokens_size: Optional[int] = 64
+    """The number of tokens to look back when applying the repeat_penalty."""
+
     @root_validator()
     def validate_environment(cls, values: Dict) -> Dict:
         """Validate that llama-cpp-python library is installed."""
@@ -92,6 +99,8 @@ class LlamaCpp(LLM, BaseModel):
         vocab_only = values["vocab_only"]
         use_mlock = values["use_mlock"]
         n_threads = values["n_threads"]
+        n_batch = values["n_batch"]
+        last_n_tokens_size = values["last_n_tokens_size"]
 
         try:
             from llama_cpp import Llama
@@ -106,6 +115,8 @@ class LlamaCpp(LLM, BaseModel):
                 vocab_only=vocab_only,
                 use_mlock=use_mlock,
                 n_threads=n_threads,
+                n_batch=n_batch,
+                last_n_tokens_size=last_n_tokens_size,
             )
         except ImportError:
             raise ModuleNotFoundError(