[Core] Support image processor (#4197)

This commit is contained in:
Cyrus Leung
2024-06-02 22:56:41 -07:00
committed by GitHub
parent dfbe60dc62
commit 7a64d24aad
29 changed files with 1042 additions and 256 deletions
+2 -23
View File
@@ -14,7 +14,6 @@ from vllm.lora.request import LoRARequest
from vllm.outputs import EmbeddingRequestOutput, RequestOutput
from vllm.pooling_params import PoolingParams
from vllm.sampling_params import SamplingParams
from vllm.sequence import MultiModalData
from vllm.usage.usage_lib import UsageContext
from vllm.utils import Counter, deprecate_kwargs
@@ -164,7 +163,6 @@ class LLM:
prompt_token_ids: Optional[List[int]] = None,
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[RequestOutput]:
...
@@ -177,7 +175,6 @@ class LLM:
prompt_token_ids: Optional[List[List[int]]] = None,
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[RequestOutput]:
...
@@ -191,7 +188,6 @@ class LLM:
prompt_token_ids: List[int],
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[RequestOutput]:
...
@@ -205,7 +201,6 @@ class LLM:
prompt_token_ids: List[List[int]],
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[RequestOutput]:
...
@@ -217,7 +212,6 @@ class LLM:
prompt_token_ids: Union[List[int], List[List[int]]],
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[RequestOutput]:
...
@@ -236,7 +230,6 @@ class LLM:
@deprecate_kwargs("prompts",
"prompt_token_ids",
"multi_modal_data",
is_deprecated=lambda: LLM.DEPRECATE_LEGACY,
additional_message="Please use the 'inputs' parameter "
"instead.")
@@ -249,7 +242,6 @@ class LLM:
prompt_token_ids: Optional[Union[List[int], List[List[int]]]] = None,
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[RequestOutput]:
"""Generates the completions for the input prompts.
@@ -281,11 +273,10 @@ class LLM:
"LLM.generate() is only supported for generation models "
"(XForCausalLM).")
if prompt_token_ids is not None or multi_modal_data is not None:
if prompt_token_ids is not None:
inputs = self._convert_v1_inputs(
prompts=cast(Optional[Union[str, List[str]]], prompts),
prompt_token_ids=prompt_token_ids,
multi_modal_data=multi_modal_data,
)
else:
inputs = cast(
@@ -314,7 +305,6 @@ class LLM:
prompt_token_ids: Optional[List[int]] = None,
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[EmbeddingRequestOutput]:
...
@@ -327,7 +317,6 @@ class LLM:
prompt_token_ids: Optional[List[List[int]]] = None,
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[EmbeddingRequestOutput]:
...
@@ -341,7 +330,6 @@ class LLM:
prompt_token_ids: List[int],
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[EmbeddingRequestOutput]:
...
@@ -355,7 +343,6 @@ class LLM:
prompt_token_ids: List[List[int]],
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[EmbeddingRequestOutput]:
...
@@ -367,7 +354,6 @@ class LLM:
prompt_token_ids: Union[List[int], List[List[int]]],
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[EmbeddingRequestOutput]:
...
@@ -386,7 +372,6 @@ class LLM:
@deprecate_kwargs("prompts",
"prompt_token_ids",
"multi_modal_data",
is_deprecated=lambda: LLM.DEPRECATE_LEGACY,
additional_message="Please use the 'inputs' parameter "
"instead.")
@@ -399,7 +384,6 @@ class LLM:
prompt_token_ids: Optional[Union[List[int], List[List[int]]]] = None,
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalData] = None,
) -> List[EmbeddingRequestOutput]:
"""Generates the completions for the input prompts.
@@ -430,11 +414,10 @@ class LLM:
"LLM.encode() is only supported for embedding models (XModel)."
)
if prompt_token_ids is not None or multi_modal_data is not None:
if prompt_token_ids is not None:
inputs = self._convert_v1_inputs(
prompts=cast(Optional[Union[str, List[str]]], prompts),
prompt_token_ids=prompt_token_ids,
multi_modal_data=multi_modal_data,
)
else:
inputs = cast(
@@ -459,7 +442,6 @@ class LLM:
self,
prompts: Optional[Union[str, List[str]]],
prompt_token_ids: Optional[Union[List[int], List[List[int]]]],
multi_modal_data: Optional[MultiModalData],
):
# skip_tokenizer_init is now checked in engine
@@ -499,9 +481,6 @@ class LLM:
else:
raise AssertionError
if multi_modal_data is not None:
item["multi_modal_data"] = multi_modal_data
inputs.append(item)
return inputs