Increased the n_batch parameter in Llama model initialization from 512 to 2048

Author: JamePeng

llama_cpp/llama.py

@@ -78,7 +78,7 @@ def __init__(
         # Context Params
         seed: int = llama_cpp.LLAMA_DEFAULT_SEED,
         n_ctx: int = 512,
-        n_batch: int = 512,
+        n_batch: int = 2048,
         n_ubatch: int = 512,
         n_seq_max: int = 1,
         n_threads: Optional[int] = None,