Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Appearance settings

Commit b9aca61

Browse filesBrowse files
abetlenabetlen
committed
misc: use typesafe byref for internal classes
1 parent a0ce429 commit b9aca61
Copy full SHA for b9aca61

File tree

Expand file treeCollapse file tree

1 file changed

+20
-20
lines changed
Filter options
Expand file treeCollapse file tree

1 file changed

+20
-20
lines changed

‎llama_cpp/_internals.py

Copy file name to clipboardExpand all lines: llama_cpp/_internals.py
+20-20Lines changed: 20 additions & 20 deletions
Original file line numberDiff line numberDiff line change
@@ -82,7 +82,7 @@ def rope_freq_scale_train(self) -> float:
8282
def desc(self) -> str:
8383
assert self.model is not None
8484
buf = ctypes.create_string_buffer(1024)
85-
llama_cpp.llama_model_desc(self.model, buf, 1024) # type: ignore
85+
llama_cpp.llama_model_desc(self.model, buf, 1024)
8686
return buf.value.decode("utf-8")
8787

8888
def size(self) -> int:
@@ -184,7 +184,7 @@ def tokenize(self, text: bytes, add_bos: bool, special: bool):
184184
def token_to_piece(self, token: int) -> bytes:
185185
assert self.model is not None
186186
buf = ctypes.create_string_buffer(32)
187-
llama_cpp.llama_token_to_piece(self.model, token, buf, 32) # type: ignore
187+
llama_cpp.llama_token_to_piece(self.model, token, buf, 32)
188188
return bytes(buf)
189189

190190
def detokenize(self, tokens: List[int]) -> bytes:
@@ -349,7 +349,7 @@ def sample_repetition_penalties(
349349
assert self.ctx is not None
350350
llama_cpp.llama_sample_repetition_penalties(
351351
self.ctx,
352-
ctypes.byref(candidates.candidates), # type: ignore
352+
llama_cpp.byref(candidates.candidates),
353353
last_tokens_data,
354354
penalty_last_n,
355355
penalty_repeat,
@@ -367,7 +367,7 @@ def sample_classifier_free_guidance(
367367
assert guidance_ctx.ctx is not None
368368
llama_cpp.llama_sample_classifier_free_guidance(
369369
self.ctx,
370-
ctypes.byref(candidates.candidates), # type: ignore
370+
llama_cpp.byref(candidates.candidates),
371371
guidance_ctx.ctx,
372372
scale,
373373
)
@@ -376,55 +376,55 @@ def sample_softmax(self, candidates: "_LlamaTokenDataArray"):
376376
assert self.ctx is not None
377377
llama_cpp.llama_sample_softmax(
378378
self.ctx,
379-
ctypes.byref(candidates.candidates), # type: ignore
379+
llama_cpp.byref(candidates.candidates),
380380
)
381381

382382
def sample_top_k(self, candidates: "_LlamaTokenDataArray", k: int, min_keep: int):
383383
assert self.ctx is not None
384384
llama_cpp.llama_sample_top_k(
385-
self.ctx, ctypes.byref(candidates.candidates), k, min_keep # type: ignore
385+
self.ctx, llama_cpp.byref(candidates.candidates), k, min_keep
386386
)
387387

388388
def sample_top_p(self, candidates: "_LlamaTokenDataArray", p: float, min_keep: int):
389389
assert self.ctx is not None
390390
llama_cpp.llama_sample_top_p(
391-
self.ctx, ctypes.byref(candidates.candidates), p, min_keep # type: ignore
391+
self.ctx, llama_cpp.byref(candidates.candidates), p, min_keep
392392
)
393393

394394
def sample_min_p(self, candidates: "_LlamaTokenDataArray", p: float, min_keep: int):
395395
assert self.ctx is not None
396396
llama_cpp.llama_sample_min_p(
397-
self.ctx, ctypes.byref(candidates.candidates), p, min_keep # type: ignore
397+
self.ctx, llama_cpp.byref(candidates.candidates), p, min_keep
398398
)
399399

400400
def sample_tail_free(
401401
self, candidates: "_LlamaTokenDataArray", z: float, min_keep: int
402402
):
403403
assert self.ctx is not None
404404
llama_cpp.llama_sample_tail_free(
405-
self.ctx, ctypes.byref(candidates.candidates), z, min_keep # type: ignore
405+
self.ctx, llama_cpp.byref(candidates.candidates), z, min_keep
406406
)
407407

408408
def sample_typical(
409409
self, candidates: "_LlamaTokenDataArray", p: float, min_keep: int
410410
):
411411
assert self.ctx is not None
412412
llama_cpp.llama_sample_typical(
413-
self.ctx, ctypes.byref(candidates.candidates), p, min_keep # type: ignore
413+
self.ctx, llama_cpp.byref(candidates.candidates), p, min_keep
414414
)
415415

416416
def sample_temp(self, candidates: "_LlamaTokenDataArray", temp: float):
417417
assert self.ctx is not None
418418
llama_cpp.llama_sample_temp(
419-
self.ctx, ctypes.byref(candidates.candidates), temp # type: ignore
419+
self.ctx, llama_cpp.byref(candidates.candidates), temp
420420
)
421421

422422
def sample_grammar(self, candidates: "_LlamaTokenDataArray", grammar: LlamaGrammar):
423423
assert self.ctx is not None
424424
assert grammar.grammar is not None
425425
llama_cpp.llama_sample_grammar(
426426
self.ctx,
427-
ctypes.byref(candidates.candidates), # type: ignore
427+
llama_cpp.byref(candidates.candidates),
428428
grammar.grammar,
429429
)
430430

@@ -434,25 +434,25 @@ def sample_token_mirostat(
434434
tau: float,
435435
eta: float,
436436
m: int,
437-
mu: ctypes._Pointer[ctypes.c_float], # type: ignore
437+
mu: llama_cpp.CtypesPointerOrRef[ctypes.c_float],
438438
) -> int:
439439
assert self.ctx is not None
440440
return llama_cpp.llama_sample_token_mirostat(
441441
self.ctx,
442-
ctypes.byref(candidates.candidates), # type: ignore
442+
llama_cpp.byref(candidates.candidates),
443443
tau,
444444
eta,
445445
m,
446446
mu,
447447
)
448448

449449
def sample_token_mirostat_v2(
450-
self, candidates: "_LlamaTokenDataArray", tau: float, eta: float, mu: ctypes._Pointer[ctypes.c_float] # type: ignore
450+
self, candidates: "_LlamaTokenDataArray", tau: float, eta: float, mu: llama_cpp.CtypesPointerOrRef[ctypes.c_float]
451451
) -> int:
452452
assert self.ctx is not None
453453
return llama_cpp.llama_sample_token_mirostat_v2(
454454
self.ctx,
455-
ctypes.byref(candidates.candidates), # type: ignore
455+
llama_cpp.byref(candidates.candidates),
456456
tau,
457457
eta,
458458
mu,
@@ -462,14 +462,14 @@ def sample_token_greedy(self, candidates: "_LlamaTokenDataArray") -> int:
462462
assert self.ctx is not None
463463
return llama_cpp.llama_sample_token_greedy(
464464
self.ctx,
465-
ctypes.byref(candidates.candidates), # type: ignore
465+
llama_cpp.byref(candidates.candidates),
466466
)
467467

468468
def sample_token(self, candidates: "_LlamaTokenDataArray") -> int:
469469
assert self.ctx is not None
470470
return llama_cpp.llama_sample_token(
471471
self.ctx,
472-
ctypes.byref(candidates.candidates), # type: ignore
472+
llama_cpp.byref(candidates.candidates),
473473
)
474474

475475
# Grammar
@@ -566,7 +566,7 @@ def __init__(self, *, n_vocab: int):
566566
size=self.n_vocab,
567567
sorted=False,
568568
)
569-
self.default_candidates_data_id = np.arange(self.n_vocab, dtype=np.intc)
569+
self.default_candidates_data_id = np.arange(self.n_vocab, dtype=np.intc) # type: ignore
570570
self.default_candidates_data_p = np.zeros(self.n_vocab, dtype=np.single)
571571

572572
def copy_logits(self, logits: npt.NDArray[np.single]):
@@ -754,7 +754,7 @@ def sample(
754754
ctx_main.sample_repetition_penalties(
755755
token_data_array,
756756
# TODO: Only create this once
757-
(llama_cpp.llama_token * len(self.prev))(*self.prev), # type: ignore
757+
(llama_cpp.llama_token * len(self.prev))(*self.prev),
758758
self.params.penalty_last_n,
759759
self.params.penalty_repeat,
760760
self.params.penalty_freq,

0 commit comments

Comments
0 (0)
Morty Proxy This is a proxified and sanitized view of the page, visit original site.