[Chore] Adjust tokenizer import to avoid circular imports (#30601)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

benchmarks/backend_request_func.py

@@ -620,7 +620,7 @@ def get_tokenizer(
         kwargs["use_fast"] = False
     if tokenizer_mode == "mistral":
         try:
-            from vllm.tokenizers import MistralTokenizer
+            from vllm.tokenizers.mistral import MistralTokenizer
         except ImportError as e:
             raise ImportError(
                 "MistralTokenizer requires vllm package.\n"

tests/entrypoints/openai/test_serving_engine.py

@@ -10,7 +10,7 @@ import pytest
 from vllm.config import ModelConfig
 from vllm.entrypoints.openai.serving_engine import OpenAIServing
 from vllm.entrypoints.openai.serving_models import OpenAIServingModels
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 
 
 @pytest.fixture()

tests/entrypoints/test_chat_utils.py

@@ -29,7 +29,8 @@ from vllm.multimodal.utils import (
     encode_image_base64,
     encode_video_base64,
 )
-from vllm.tokenizers import MistralTokenizer, get_tokenizer
+from vllm.tokenizers import get_tokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.utils.serial_utils import tensor2base64
 
 from ..models.registry import HF_EXAMPLE_MODELS

tests/models/language/generation/test_mistral.py

@@ -10,7 +10,7 @@ from vllm.entrypoints.openai.tool_parsers.mistral_tool_parser import (
     MistralToolParser,
 )
 from vllm.sampling_params import SamplingParams
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 
 from ...utils import check_logprobs_close
 

tests/models/multimodal/generation/test_voxtral.py

@@ -9,7 +9,7 @@ from mistral_common.audio import Audio
 from mistral_common.protocol.instruct.chunk import AudioChunk, RawAudio, TextChunk
 from mistral_common.protocol.instruct.messages import UserMessage
 
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 
 from ....conftest import AudioTestAssets
 from ....utils import RemoteOpenAIServer

tests/models/multimodal/processing/test_common.py

@@ -22,11 +22,8 @@ from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalDataDict
 from vllm.multimodal.cache import MultiModalProcessorOnlyCache
 from vllm.multimodal.inputs import MultiModalInputs, batched_tensors_equal
 from vllm.multimodal.processing import BaseMultiModalProcessor, InputProcessingContext
-from vllm.tokenizers import (
+from vllm.tokenizers import TokenizerLike, cached_tokenizer_from_config
-    MistralTokenizer,
+from vllm.tokenizers.mistral import MistralTokenizer
-    TokenizerLike,
-    cached_tokenizer_from_config,
-)
 
 from ....multimodal.utils import random_audio, random_image, random_video
 from ...registry import (

tests/reasoning/test_mistral_reasoning_parser.py

@@ -5,7 +5,7 @@ import pytest
 
 from tests.reasoning.utils import run_reasoning_extraction_mistral
 from vllm.reasoning import ReasoningParser, ReasoningParserManager
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 
 parser_name = "mistral"
 

tests/reasoning/utils.py

@@ -4,7 +4,7 @@
 
 from vllm.entrypoints.openai.protocol import ChatCompletionRequest, DeltaMessage
 from vllm.reasoning import ReasoningParser
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 
 
 class StreamingReasoningReconstructor:

tests/tokenizers_/test_detokenize.py

@@ -8,7 +8,7 @@ import pytest
 from transformers import AutoTokenizer, PreTrainedTokenizer, PreTrainedTokenizerFast
 
 from vllm.sampling_params import SamplingParams
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.v1.engine import EngineCoreRequest
 from vllm.v1.engine.detokenizer import (
     FastIncrementalDetokenizer,

tests/tool_use/test_mistral_tool_parser.py

@@ -13,12 +13,9 @@ from partial_json_parser.core.options import Allow
 
 from vllm.entrypoints.openai.protocol import DeltaMessage, DeltaToolCall
 from vllm.entrypoints.openai.tool_parsers.mistral_tool_parser import MistralToolParser
-from vllm.tokenizers import (
+from vllm.tokenizers import TokenizerLike, get_tokenizer
-    MistralTokenizer,
-    TokenizerLike,
-    get_tokenizer,
-)
 from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
+from vllm.tokenizers.mistral import MistralTokenizer
 
 
 @pytest.fixture(scope="module")

vllm/entrypoints/chat_utils.py

@@ -49,7 +49,8 @@ from vllm.logger import init_logger
 from vllm.model_executor.models import SupportsMultiModal
 from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalDataDict, MultiModalUUIDDict
 from vllm.multimodal.utils import MEDIA_CONNECTOR_REGISTRY, MediaConnector
-from vllm.tokenizers import MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.transformers_utils.chat_templates import get_chat_template_fallback_path
 from vllm.transformers_utils.processor import cached_get_processor
 from vllm.utils import random_uuid

vllm/entrypoints/llm.py

@@ -72,7 +72,8 @@ from vllm.platforms import current_platform
 from vllm.pooling_params import PoolingParams
 from vllm.sampling_params import BeamSearchParams, RequestOutputKind, SamplingParams
 from vllm.tasks import PoolingTask
-from vllm.tokenizers import MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.usage.usage_lib import UsageContext
 from vllm.utils.collection_utils import as_iter, is_list_of
 from vllm.utils.counter import Counter

vllm/entrypoints/openai/serving_engine.py

@@ -117,7 +117,9 @@ from vllm.outputs import CompletionOutput, PoolingRequestOutput, RequestOutput
 from vllm.pooling_params import PoolingParams
 from vllm.reasoning import ReasoningParser, ReasoningParserManager
 from vllm.sampling_params import BeamSearchParams, SamplingParams
-from vllm.tokenizers import DeepseekV32Tokenizer, MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.deepseekv32 import DeepseekV32Tokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.tracing import (
     contains_trace_headers,
     extract_trace_headers,

vllm/entrypoints/openai/tool_parsers/hermes_tool_parser.py

@@ -22,7 +22,8 @@ from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
 from vllm.logger import init_logger
-from vllm.tokenizers import MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.mistral import MistralTokenizer
 
 logger = init_logger(__name__)
 

vllm/entrypoints/openai/tool_parsers/jamba_tool_parser.py

@@ -21,7 +21,8 @@ from vllm.entrypoints.openai.protocol import (
 from vllm.entrypoints.openai.tool_parsers import ToolParser
 from vllm.entrypoints.openai.tool_parsers.utils import extract_intermediate_diff
 from vllm.logger import init_logger
-from vllm.tokenizers import MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.mistral import MistralTokenizer
 
 logger = init_logger(__name__)
 

vllm/entrypoints/openai/tool_parsers/mistral_tool_parser.py

@@ -6,6 +6,7 @@ from collections.abc import Sequence
 from enum import Enum, auto
 from random import choices
 from string import ascii_letters, digits
+from typing import Any
 
 import ijson
 import regex as re
@@ -24,7 +25,8 @@ from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
 from vllm.logger import init_logger
-from vllm.tokenizers import MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.mistral import MistralTokenizer
 
 logger = init_logger(__name__)
 
@@ -84,6 +86,7 @@ class MistralToolParser(ToolParser):
 
         # initialize properties used for state when parsing tool calls in
         # streaming mode
+        self.prev_tool_call_arr: list[dict[str, Any]] = []
         self.current_tool_id: int = -1
         self.streaming_state: StreamingState = StreamingState.WAITING_FOR_TOOL_START
 

vllm/entrypoints/pooling/score/serving.py

@@ -38,7 +38,8 @@ from vllm.inputs.data import TokensPrompt
 from vllm.logger import init_logger
 from vllm.lora.request import LoRARequest
 from vllm.outputs import PoolingRequestOutput, ScoringRequestOutput
-from vllm.tokenizers import MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.utils.async_utils import make_async, merge_async_iterators
 
 logger = init_logger(__name__)

vllm/entrypoints/utils.py

@@ -30,7 +30,7 @@ from vllm.entrypoints.openai.protocol import (
 from vllm.entrypoints.openai.serving_models import LoRAModulePath
 from vllm.logger import init_logger
 from vllm.platforms import current_platform
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 logger = init_logger(__name__)

vllm/model_executor/models/pixtral.py

@@ -59,7 +59,8 @@ from vllm.multimodal.processing import (
 from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
 from vllm.platforms import current_platform
 from vllm.sequence import IntermediateTensors
-from vllm.tokenizers import MistralTokenizer, cached_tokenizer_from_config
+from vllm.tokenizers import cached_tokenizer_from_config
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.utils.tensor_schema import TensorSchema, TensorShape
 
 from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsPP

vllm/model_executor/models/voxtral.py

@@ -51,7 +51,8 @@ from vllm.multimodal.processing import (
 )
 from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
 from vllm.sequence import IntermediateTensors
-from vllm.tokenizers import MistralTokenizer, cached_tokenizer_from_config
+from vllm.tokenizers import cached_tokenizer_from_config
+from vllm.tokenizers.mistral import MistralTokenizer
 
 from .interfaces import SupportsLoRA, SupportsMultiModal, SupportsTranscription
 from .utils import init_vllm_registered_model, maybe_prefix

vllm/reasoning/mistral_reasoning_parser.py

@@ -10,7 +10,7 @@ from vllm.entrypoints.openai.protocol import (
 from vllm.logger import init_logger
 from vllm.reasoning import ReasoningParser
 from vllm.reasoning.basic_parsers import BaseThinkingReasoningParser
-from vllm.tokenizers import MistralTokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 
 logger = init_logger(__name__)
 

vllm/v1/engine/input_processor.py

@@ -19,7 +19,8 @@ from vllm.multimodal.processing import EncDecMultiModalProcessor
 from vllm.multimodal.utils import argsort_mm_positions
 from vllm.pooling_params import PoolingParams
 from vllm.sampling_params import SamplingParams
-from vllm.tokenizers import MistralTokenizer, TokenizerLike
+from vllm.tokenizers import TokenizerLike
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.utils import length_from_prompt_token_ids_or_embeds
 from vllm.v1.engine import EngineCoreRequest
 from vllm.v1.metrics.stats import MultiModalCacheStats

vllm/v1/structured_output/backend_xgrammar.py

@@ -10,7 +10,8 @@ import torch
 import vllm.envs
 from vllm.logger import init_logger
 from vllm.sampling_params import SamplingParams
-from vllm.tokenizers import DeepseekV32Tokenizer, MistralTokenizer
+from vllm.tokenizers.deepseekv32 import DeepseekV32Tokenizer
+from vllm.tokenizers.mistral import MistralTokenizer
 from vllm.utils.import_utils import LazyLoader
 from vllm.v1.structured_output.backend_types import (
     StructuredOutputBackend,