[Chore] Remove redundant input parsing methods (#33542)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
@@ -68,7 +68,6 @@ from vllm.entrypoints.openai.parser.harmony_utils import (
|
||||
from vllm.entrypoints.openai.utils import maybe_filter_parallel_tool_calls
|
||||
from vllm.entrypoints.utils import get_max_tokens, should_include_usage
|
||||
from vllm.inputs.data import EmbedsPrompt, TokensPrompt
|
||||
from vllm.inputs.parse import get_prompt_components
|
||||
from vllm.logger import init_logger
|
||||
from vllm.logprobs import Logprob
|
||||
from vllm.outputs import CompletionOutput, RequestOutput
|
||||
@@ -359,7 +358,7 @@ class OpenAIServingChat(OpenAIServing):
|
||||
generators: list[AsyncGenerator[RequestOutput, None]] = []
|
||||
try:
|
||||
for i, engine_prompt in enumerate(engine_prompts):
|
||||
prompt_text, _, _ = get_prompt_components(engine_prompt)
|
||||
prompt_text = engine_prompt.get("prompt")
|
||||
|
||||
# If we are creating sub requests for multiple prompts, ensure that they
|
||||
# have unique request ids.
|
||||
|
||||
@@ -34,8 +34,7 @@ from vllm.entrypoints.openai.engine.serving import (
|
||||
from vllm.entrypoints.openai.models.serving import OpenAIServingModels
|
||||
from vllm.entrypoints.utils import get_max_tokens, should_include_usage
|
||||
from vllm.exceptions import VLLMValidationError
|
||||
from vllm.inputs.data import EmbedsPrompt, TokensPrompt, is_embeds_prompt
|
||||
from vllm.inputs.parse import get_prompt_components
|
||||
from vllm.inputs.data import EmbedsPrompt, TokensPrompt
|
||||
from vllm.logger import init_logger
|
||||
from vllm.logprobs import Logprob
|
||||
from vllm.outputs import RequestOutput
|
||||
@@ -161,7 +160,7 @@ class OpenAIServingCompletion(OpenAIServing):
|
||||
generators: list[AsyncGenerator[RequestOutput, None]] = []
|
||||
try:
|
||||
for i, engine_prompt in enumerate(engine_prompts):
|
||||
prompt_text, _, _ = get_prompt_components(engine_prompt)
|
||||
prompt_text = engine_prompt.get("prompt")
|
||||
|
||||
max_tokens = get_max_tokens(
|
||||
max_model_len=self.max_model_len,
|
||||
@@ -278,11 +277,7 @@ class OpenAIServingCompletion(OpenAIServing):
|
||||
# with the inputs token IDs
|
||||
if final_res.prompt is None:
|
||||
engine_prompt = engine_prompts[i]
|
||||
final_res.prompt = (
|
||||
None
|
||||
if is_embeds_prompt(engine_prompt)
|
||||
else engine_prompt.get("prompt")
|
||||
)
|
||||
final_res.prompt = engine_prompt.get("prompt")
|
||||
|
||||
final_res_batch_checked = cast(list[RequestOutput], final_res_batch)
|
||||
|
||||
@@ -352,11 +347,7 @@ class OpenAIServingCompletion(OpenAIServing):
|
||||
prompt_text = res.prompt
|
||||
if prompt_text is None:
|
||||
engine_prompt = engine_prompts[prompt_idx]
|
||||
prompt_text = (
|
||||
None
|
||||
if is_embeds_prompt(engine_prompt)
|
||||
else engine_prompt.get("prompt")
|
||||
)
|
||||
prompt_text = engine_prompt.get("prompt")
|
||||
|
||||
# Prompt details are excluded from later streamed outputs
|
||||
if prompt_token_ids is not None:
|
||||
|
||||
@@ -1116,7 +1116,7 @@ class OpenAIServing:
|
||||
priority: int = 0,
|
||||
trace_headers: Mapping[str, str] | None = None,
|
||||
):
|
||||
prompt_text, _, _ = get_prompt_components(engine_prompt)
|
||||
prompt_text = engine_prompt.get("prompt")
|
||||
|
||||
orig_priority = priority
|
||||
sub_request = 0
|
||||
@@ -1186,7 +1186,7 @@ class OpenAIServing:
|
||||
context.chat_template_content_format,
|
||||
)
|
||||
engine_prompt = engine_prompts[0]
|
||||
prompt_text, _, _ = get_prompt_components(engine_prompt)
|
||||
prompt_text = engine_prompt.get("prompt")
|
||||
|
||||
sampling_params.max_tokens = get_max_tokens(
|
||||
self.max_model_len,
|
||||
|
||||
Reference in New Issue
Block a user