@@ -108,7 +108,7 @@ def apply_lora_from_file(
108108 scale ,
109109 path_base_model .encode ("utf-8" )
110110 if path_base_model is not None
111- else llama_cpp .c_char_p (0 ),
111+ else ctypes .c_char_p (0 ),
112112 n_threads ,
113113 )
114114
@@ -303,8 +303,8 @@ def decode(self, batch: "_LlamaBatch"):
303303 assert self .ctx is not None
304304 assert batch .batch is not None
305305 return_code = llama_cpp .llama_decode (
306- ctx = self .ctx ,
307- batch = batch .batch ,
306+ self .ctx ,
307+ batch .batch ,
308308 )
309309 if return_code != 0 :
310310 raise RuntimeError (f"llama_decode returned { return_code } )
@@ -493,7 +493,7 @@ class _LlamaBatch:
493493 def __init__ (
494494 self , * , n_tokens : int , embd : int , n_seq_max : int , verbose : bool = True
495495 ):
496- self .n_tokens = n_tokens
496+ self ._n_tokens = n_tokens
497497 self .embd = embd
498498 self .n_seq_max = n_seq_max
499499 self .verbose = verbose
@@ -502,7 +502,7 @@ def __init__(
502502
503503 self .batch = None
504504 self .batch = llama_cpp .llama_batch_init (
505- self .n_tokens , self .embd , self .n_seq_max
505+ self ._n_tokens , self .embd , self .n_seq_max
506506 )
507507
508508 def __del__ (self ):
@@ -570,12 +570,13 @@ def copy_logits(self, logits: npt.NDArray[np.single]):
570570 self .candidates .data = self .candidates_data .ctypes .data_as (
571571 llama_cpp .llama_token_data_p
572572 )
573- self .candidates .sorted = llama_cpp .c_bool (False )
574- self .candidates .size = llama_cpp .c_size_t (self .n_vocab )
573+ self .candidates .sorted = ctypes .c_bool (False )
574+ self .candidates .size = ctypes .c_size_t (self .n_vocab )
575575
576576
577577# Python wrappers over common/common
578578def _tokenize (model : _LlamaModel , text : str , add_bos : bool , special : bool ) -> list [int ]:
579+ assert model .model is not None
579580 n_tokens = len (text ) + 1 if add_bos else len (text )
580581 result = (llama_cpp .llama_token * n_tokens )()
581582 n_tokens = llama_cpp .llama_tokenize (
0 commit comments