Remove Python 3.9 support (#145)

* Remove Python 3.9 support

* Fix typing issue

* Fix

* Fix issue from merge

Author: juliendenize

.github/workflows/lint_build_test.yaml

@@ -27,7 +27,6 @@ jobs:
       matrix:
         # Support all LTS versions of Python
         python-version: [
-          "3.9",
           "3.10",
           "3.11",
           "3.12",

pyproject.toml

@@ -10,7 +10,7 @@ authors = [{name = "bam4d", email = "[email protected]"}]
 license = "Apache-2.0"
 license-files = ["LICENSE"] 
 readme = "README.md"
-requires-python = ">=3.9.0,<3.14"
+requires-python = ">=3.10.0,<3.14"
 
 dependencies = [
     "pydantic>=2.7,<3.0",

src/mistral_common/audio.py

@@ -5,7 +5,7 @@
 from enum import Enum
 from functools import cache
 from pathlib import Path
-from typing import TYPE_CHECKING, Union
+from typing import TYPE_CHECKING
 
 import numpy as np
 import requests
@@ -228,7 +228,7 @@ def resample(self, new_sampling_rate: int) -> None:
         self.sampling_rate = new_sampling_rate
 
 
-def hertz_to_mel(freq: Union[float, np.ndarray]) -> Union[float, np.ndarray]:
+def hertz_to_mel(freq: float | np.ndarray) -> float | np.ndarray:
     r"""Convert frequency from hertz to mels using the "slaney" mel-scale.
 
     Args:

src/mistral_common/exceptions.py

@@ -1,6 +1,3 @@
-from typing import Optional
-
-
 class MistralCommonException(Exception):
     r"""Base class for all Mistral exceptions.
 
@@ -12,7 +9,7 @@ class MistralCommonException(Exception):
 
     def __init__(
         self,
-        message: Optional[str] = None,
+        message: str | None = None,
     ) -> None:
         r"""Initialize the `MistralCommonException` with an optional message.
 

src/mistral_common/experimental/app/main.py

@@ -1,6 +1,5 @@
 from functools import lru_cache
 from pathlib import Path
-from typing import Union
 
 import click
 import uvicorn
@@ -21,7 +20,7 @@
 
 
 def create_app(
-    tokenizer: Union[str, Path, MistralTokenizer],
+    tokenizer: str | Path | MistralTokenizer,
     validation_mode: ValidationMode = ValidationMode.test,
     engine_url: str = "127.0.0.1",
     engine_backend: EngineBackend = EngineBackend.llama_cpp,
@@ -114,8 +113,8 @@ def cli() -> None:
     show_default=True,
 )
 def serve(
-    tokenizer_path: Union[str, Path],
-    validation_mode: Union[ValidationMode, str] = ValidationMode.test,
+    tokenizer_path: str | Path,
+    validation_mode: ValidationMode | str = ValidationMode.test,
     host: str = "127.0.0.1",
     port: int = 0,
     engine_url: str = "http://127.0.0.1:8080",

src/mistral_common/experimental/app/models.py

@@ -1,7 +1,7 @@
 import importlib.metadata
 from enum import Enum
 from pathlib import Path
-from typing import Any, List, Optional, Union
+from typing import Any
 
 from pydantic import BaseModel, ConfigDict, field_validator
 from pydantic_settings import BaseSettings
@@ -23,8 +23,8 @@ class OpenAIChatCompletionRequest(BaseModel):
         This class accepts extra fields that are not validated.
     """
 
-    messages: List[dict[str, Union[str, List[dict[str, Union[str, dict[str, Any]]]]]]]
-    tools: Optional[List[dict[str, Any]]] = None
+    messages: list[dict[str, str | list[dict[str, str | dict[str, Any]]]]]
+    tools: list[dict[str, Any]] | None = None
 
     # Allow extra fields as the `from_openai` method will handle them.
     # We never validate the input, so we don't need to worry about the extra fields.
@@ -83,16 +83,16 @@ def _validate_engine_url(cls, value: str) -> str:
 
     @field_validator("engine_backend", mode="before")
     @classmethod
-    def _validate_backend(cls, value: Union[str, EngineBackend]) -> EngineBackend:
+    def _validate_backend(cls, value: str | EngineBackend) -> EngineBackend:
         if isinstance(value, str):
             value = EngineBackend(value)
         return value
 
     def model_post_init(self, context: Any) -> None:
         super().model_post_init(context)
-        self._tokenizer: Optional[MistralTokenizer] = None
+        self._tokenizer: MistralTokenizer | None = None
 
-    def _load_tokenizer(self, tokenizer_path: Union[str, Path], validation_mode: ValidationMode) -> None:
+    def _load_tokenizer(self, tokenizer_path: str | Path, validation_mode: ValidationMode) -> None:
         if tokenizer_path == "":
             raise ValueError("Tokenizer path must be set via the environment variable `TOKENIZER_PATH`.")
         elif self._tokenizer is not None:

src/mistral_common/experimental/app/routers.py

@@ -1,5 +1,5 @@
 import json
-from typing import Annotated, List, Optional, Union
+from typing import Annotated
 
 import requests
 from fastapi import APIRouter, Body, Depends, HTTPException
@@ -33,7 +33,7 @@ async def redirect_to_docs() -> RedirectResponse:
 
 @tokenize_router.post("/")
 async def tokenize_request(
-    request: Union[ChatCompletionRequest, OpenAIChatCompletionRequest],
+    request: ChatCompletionRequest | OpenAIChatCompletionRequest,
     settings: Annotated[Settings, Depends(get_settings)],
 ) -> list[int]:
     r"""Tokenize a chat completion request."""
@@ -100,7 +100,7 @@ async def detokenize_to_assistant_message(
     else:
         content_tokens, tool_calls_tokens = tokens, ()
 
-    content: Optional[Union[str, List[Union[TextChunk, ThinkChunk]]]] = None
+    content: str | list[TextChunk | ThinkChunk] | None = None
 
     if settings.tokenizer.instruct_tokenizer.tokenizer.version >= TokenizerVersion.v13:
         assert isinstance(settings.tokenizer.instruct_tokenizer, InstructTokenizerV13)
@@ -150,7 +150,7 @@ async def detokenize_to_assistant_message(
 
 @main_router.post("/v1/chat/completions", tags=["chat", "completions"])
 async def generate(
-    request: Union[ChatCompletionRequest, OpenAIChatCompletionRequest],
+    request: ChatCompletionRequest | OpenAIChatCompletionRequest,
     settings: Annotated[Settings, Depends(get_settings)],
 ) -> AssistantMessage:
     r"""Generate a chat completion.

src/mistral_common/experimental/think.py

@@ -1,25 +1,22 @@
-from typing import List, Tuple
-
-
 def _split_content_and_think_chunks(
-    tokens: List[int], begin_think_token_id: int, end_think_token_id: int
-) -> List[Tuple[List[int], bool]]:
+    tokens: list[int], begin_think_token_id: int, end_think_token_id: int
+) -> list[tuple[list[int], bool]]:
     r"""Split the content and think chunks from a list of tokens.
 
     Args:
-        tokens: List of tokens.
+        tokens: list of tokens.
         begin_think_token_id: The token id for the begin think token.
         end_think_token_id: The token id for the end think token.
 
     Returns:
-        List of tuples, where each tuple contains a list of tokens and a boolean indicating if the chunk is a think
+        list of tuples, where each tuple contains a list of tokens and a boolean indicating if the chunk is a think
         chunk.
     """
     if not tokens:
         return []
 
-    content_chunks: List[Tuple[List[int], bool]] = []
-    current_content: List[int] = []
+    content_chunks: list[tuple[list[int], bool]] = []
+    current_content: list[int] = []
 
     in_think_chunk = False
     for token in tokens:

src/mistral_common/experimental/tools.py

@@ -1,5 +1,5 @@
 import json
-from typing import List, Sequence, Tuple
+from typing import Sequence
 
 from mistral_common.experimental.utils import (
     _split_integer_list_by_value,
@@ -18,8 +18,8 @@ class InvalidArgsToolCallError(InvalidToolCallError):
 
 
 def _split_content_and_tool_calls(
-    tokens: List[int], tool_call_token_id: int
-) -> tuple[List[int], Tuple[List[int], ...]]:
+    tokens: list[int], tool_call_token_id: int
+) -> tuple[list[int], tuple[list[int], ...]]:
     r"""Split the content and tool calls from a list of tokens.
 
     The content is the first sequence of tokens that does not start with the tool call token ID.
@@ -48,7 +48,7 @@ def _split_content_and_tool_calls(
     return content_tokens, tools_calls_tokens
 
 
-def _decode_tool_calls_v2_up_to_v7(tool_call_tokens: list[int], tokenizer: Tokenizer) -> List[ToolCall]:
+def _decode_tool_calls_v2_up_to_v7(tool_call_tokens: list[int], tokenizer: Tokenizer) -> list[ToolCall]:
     r"""Decode a list of tool call tokens into a list of tool calls for tokenizer versions v2 to v7.
 
     Note:
@@ -130,7 +130,7 @@ def _decode_tool_call_v11(tool_call_tokens: list[int], tokenizer: Tokenizer) ->
     return tool_call
 
 
-def _decode_tool_calls(tool_call_tokens: Sequence[list[int]], tokenizer: Tokenizer) -> List[ToolCall]:
+def _decode_tool_calls(tool_call_tokens: Sequence[list[int]], tokenizer: Tokenizer) -> list[ToolCall]:
     r"""Decode a list of tool call tokens into a list of tool calls.
 
     Note:

src/mistral_common/experimental/utils.py

@@ -1,9 +1,7 @@
-from typing import List, Tuple
-
 from mistral_common.tokens.tokenizers.base import Tokenizer
 
 
-def _split_integer_list_by_value(list_: List[int], value: int) -> Tuple[List[int], ...]:
+def _split_integer_list_by_value(list_: list[int], value: int) -> tuple[list[int], ...]:
     r"""Split a list of integers by a given value.
 
     Args:
@@ -32,8 +30,8 @@ def _split_integer_list_by_value(list_: List[int], value: int) -> Tuple[List[int
 
 
 def _split_tokens_by_one_occurence_control_token(
-    list_: List[int], tokenizer: Tokenizer, control_token: str
-) -> Tuple[List[int], List[int]]:
+    list_: list[int], tokenizer: Tokenizer, control_token: str
+) -> tuple[list[int], list[int]]:
     r"""Split a list of integers by a given control token.
 
     Raises: