A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 16, 05:45 UTC
[Bugfix] Fix usage of `deprecated` decorator (#11025)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Dec 10, 2024 at 13:45 UTC
391d7b2763df0b90a975c7232f38c4de4be2ff85
4 files changed
+52
-52
vllm/engine/llm_engine.py
+4
-4
@@ -677,12 +677,10 @@ class LLMEngine:
677
self.model_executor.stop_remote_worker_execution_loop()
678
679
@overload
680
- @deprecated("'inputs' will be renamed to 'prompt")
680
def add_request(
681
self,
682
request_id: str,
684
- *,
685
- inputs: PromptType,
683
+ prompt: PromptType,
684
params: Union[SamplingParams, PoolingParams],
685
arrival_time: Optional[float] = None,
686
lora_request: Optional[LoRARequest] = None,
@@ -693,10 +691,12 @@ class LLMEngine:
691
...
692
693
@overload
694
+ @deprecated("'inputs' will be renamed to 'prompt")
695
def add_request(
696
self,
697
request_id: str,
699
- prompt: PromptType,
698
+ *,
699
+ inputs: PromptType,
700
params: Union[SamplingParams, PoolingParams],
701
arrival_time: Optional[float] = None,
702
lora_request: Optional[LoRARequest] = None,
vllm/engine/multiprocessing/__init__.py
+4
-4
@@ -35,11 +35,9 @@ class RPCProcessRequest:
35
priority: int = 0
36
37
@overload
38
- @deprecated("'inputs' will be renamed to 'prompt")
38
def __init__(
39
self,
41
- *,
42
- inputs: PromptType,
40
+ prompt: PromptType,
41
params: Union[SamplingParams, PoolingParams],
42
request_id: str,
43
lora_request: Optional[LoRARequest] = None,
@@ -50,9 +48,11 @@ class RPCProcessRequest:
48
...
49
50
@overload
51
+ @deprecated("'inputs' will be renamed to 'prompt")
52
def __init__(
53
self,
55
- prompt: PromptType,
54
+ *,
55
+ inputs: PromptType,
56
params: Union[SamplingParams, PoolingParams],
57
request_id: str,
58
lora_request: Optional[LoRARequest] = None,
vllm/engine/multiprocessing/client.py
+8
-8
@@ -415,11 +415,9 @@ class MQLLMEngineClient(EngineClient):
415
return ENGINE_DEAD_ERROR(self._errored_with)
416
417
@overload
418
- @deprecated("'inputs' will be renamed to 'prompt")
418
def generate(
419
self,
421
- *,
422
- inputs: PromptType,
420
+ prompt: PromptType,
421
sampling_params: SamplingParams,
422
request_id: str,
423
lora_request: Optional[LoRARequest] = None,
@@ -430,9 +428,11 @@ class MQLLMEngineClient(EngineClient):
428
...
429
430
@overload
431
+ @deprecated("'inputs' will be renamed to 'prompt")
432
def generate(
433
self,
435
- prompt: PromptType,
434
+ *,
435
+ inputs: PromptType,
436
sampling_params: SamplingParams,
437
request_id: str,
438
lora_request: Optional[LoRARequest] = None,
@@ -487,11 +487,9 @@ class MQLLMEngineClient(EngineClient):
487
prompt_adapter_request, priority)
488
489
@overload
490
- @deprecated("'inputs' will be renamed to 'prompt")
490
def encode(
491
self,
493
- *,
494
- inputs: PromptType,
492
+ prompt: PromptType,
493
pooling_params: PoolingParams,
494
request_id: str,
495
lora_request: Optional[LoRARequest] = None,
@@ -501,9 +499,11 @@ class MQLLMEngineClient(EngineClient):
499
...
500
501
@overload
502
+ @deprecated("'inputs' will be renamed to 'prompt")
503
def encode(
504
self,
506
- prompt: PromptType,
505
+ *,
506
+ inputs: PromptType,
507
pooling_params: PoolingParams,
508
request_id: str,
509
lora_request: Optional[LoRARequest] = None,
vllm/entrypoints/llm.py
+36
-36
@@ -252,8 +252,21 @@ class LLM:
252
else:
253
tokenizer_group.tokenizer = get_cached_tokenizer(tokenizer)
254
255
+ @overload
256
+ def generate(
257
+ self,
258
+ prompts: Union[PromptType, Sequence[PromptType]],
259
+ /,
260
+ *,
261
+ sampling_params: Optional[Union[SamplingParams,
262
+ Sequence[SamplingParams]]] = None,
263
+ use_tqdm: bool = True,
264
+ lora_request: Optional[Union[List[LoRARequest], LoRARequest]] = None,
265
+ ) -> List[RequestOutput]:
266
+ ...
267
+
268
@overload # LEGACY: single (prompt + optional token ids)
256
- @deprecated("'prompt_token_ids' will become part of 'prompts")
269
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
270
def generate(
271
self,
272
prompts: str,
@@ -266,7 +279,7 @@ class LLM:
279
...
280
281
@overload # LEGACY: multi (prompt + optional token ids)
269
- @deprecated("'prompt_token_ids' will become part of 'prompts")
282
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
283
def generate(
284
self,
285
prompts: List[str],
@@ -279,7 +292,7 @@ class LLM:
292
...
293
294
@overload # LEGACY: single (token ids + optional prompt)
282
- @deprecated("'prompt_token_ids' will become part of 'prompts")
295
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
296
def generate(
297
self,
298
prompts: Optional[str] = None,
@@ -293,7 +306,7 @@ class LLM:
306
...
307
308
@overload # LEGACY: multi (token ids + optional prompt)
296
- @deprecated("'prompt_token_ids' will become part of 'prompts")
309
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
310
def generate(
311
self,
312
prompts: Optional[List[str]] = None,
@@ -307,7 +320,7 @@ class LLM:
320
...
321
322
@overload # LEGACY: single or multi token ids [pos-only]
310
- @deprecated("'prompt_token_ids' will become part of 'prompts")
323
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
324
def generate(
325
self,
326
prompts: None,
@@ -318,19 +331,6 @@ class LLM:
331
) -> List[RequestOutput]:
332
...
333
321
- @overload
322
- def generate(
323
- self,
324
- prompts: Union[PromptType, Sequence[PromptType]],
325
- /,
326
- *,
327
- sampling_params: Optional[Union[SamplingParams,
328
- Sequence[SamplingParams]]] = None,
329
- use_tqdm: bool = True,
330
- lora_request: Optional[Union[List[LoRARequest], LoRARequest]] = None,
331
- ) -> List[RequestOutput]:
332
- ...
333
-
334
@deprecate_kwargs(
335
"prompt_token_ids",
336
is_deprecated=lambda: LLM.DEPRECATE_LEGACY,
@@ -672,8 +672,21 @@ class LLM:
672
lora_request=lora_request,
673
)
674
675
+ @overload
676
+ def encode(
677
+ self,
678
+ prompts: Union[PromptType, Sequence[PromptType]],
679
+ /,
680
+ *,
681
+ pooling_params: Optional[Union[PoolingParams,
682
+ Sequence[PoolingParams]]] = None,
683
+ use_tqdm: bool = True,
684
+ lora_request: Optional[Union[List[LoRARequest], LoRARequest]] = None,
685
+ ) -> List[PoolingRequestOutput]:
686
+ ...
687
+
688
@overload # LEGACY: single (prompt + optional token ids)
676
- @deprecated("'prompt_token_ids' will become part of 'prompts")
689
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
690
def encode(
691
self,
692
prompts: str,
@@ -686,7 +699,7 @@ class LLM:
699
...
700
701
@overload # LEGACY: multi (prompt + optional token ids)
689
- @deprecated("'prompt_token_ids' will become part of 'prompts")
702
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
703
def encode(
704
self,
705
prompts: List[str],
@@ -699,7 +712,7 @@ class LLM:
712
...
713
714
@overload # LEGACY: single (token ids + optional prompt)
702
- @deprecated("'prompt_token_ids' will become part of 'prompts")
715
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
716
def encode(
717
self,
718
prompts: Optional[str] = None,
@@ -713,7 +726,7 @@ class LLM:
726
...
727
728
@overload # LEGACY: multi (token ids + optional prompt)
716
- @deprecated("'prompt_token_ids' will become part of 'prompts")
729
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
730
def encode(
731
self,
732
prompts: Optional[List[str]] = None,
@@ -727,7 +740,7 @@ class LLM:
740
...
741
742
@overload # LEGACY: single or multi token ids [pos-only]
730
- @deprecated("'prompt_token_ids' will become part of 'prompts")
743
+ @deprecated("'prompt_token_ids' will become part of 'prompts'")
744
def encode(
745
self,
746
prompts: None,
@@ -738,19 +751,6 @@ class LLM:
751
) -> List[PoolingRequestOutput]:
752
...
753
741
- @overload
742
- def encode(
743
- self,
744
- prompts: Union[PromptType, Sequence[PromptType]],
745
- /,
746
- *,
747
- pooling_params: Optional[Union[PoolingParams,
748
- Sequence[PoolingParams]]] = None,
749
- use_tqdm: bool = True,
750
- lora_request: Optional[Union[List[LoRARequest], LoRARequest]] = None,
751
- ) -> List[PoolingRequestOutput]:
752
- ...
753
-
754
@deprecate_kwargs(
755
"prompt_token_ids",
756
is_deprecated=lambda: LLM.DEPRECATE_LEGACY,