poetry unlock (#430)

* poetry unlock

* remove embed lock

libs/embed_package/Makefile

@@ -32,13 +32,12 @@ lint format: PYTHON_FILES=.
 lint_diff format_diff: PYTHON_FILES=$(shell git diff --relative=libs/infinity_emb --name-only --diff-filter=d master | grep -E '\.py$$|\.ipynb$$')
 
 lint lint_diff:
-	poetry run ruff .
-	[ "$(PYTHON_FILES)" = "" ] || poetry run black $(PYTHON_FILES) --check
+	poetry run ruff check .
+
 	[ "$(PYTHON_FILES)" = "" ] || poetry run mypy $(PYTHON_FILES)
 
 format format_diff:
-	[ "$(PYTHON_FILES)" = "" ] || poetry run black $(PYTHON_FILES)
-	[ "$(PYTHON_FILES)" = "" ] || poetry run ruff --select I --fix $(PYTHON_FILES)
+	[ "$(PYTHON_FILES)" = "" ] || poetry run ruff format $(PYTHON_FILES)
 
 poetry_check:
 	poetry check

libs/embed_package/pyproject.toml

@@ -18,9 +18,9 @@ python = ">=3.9,<4"
 # infinity_emb = {path = "../infinity_emb", extras = ["optimum","vision","torch","audio"]}
 infinity_emb = {version = "0.0.58", extras = ["optimum","vision","torch","audio"]}
 [tool.poetry.group.test.dependencies]
-pytest = "^7.0.0"
-coverage = {extras = ["toml"], version = "^7.3.2"}
-mypy = "^1.5.1"
+pytest = "^8.0.0"
+coverage = {extras = ["toml"], version = "^7.6.3"}
+mypy = "^1.12.0"
 
 [tool.poetry.group.codespell.dependencies]
 codespell = "^2.2.0"

libs/infinity_emb/Makefile

@@ -30,17 +30,15 @@ openapi:
 
 # Define a variable for Python and notebook files.
 PYTHON_FILES=.
-lint format: PYTHON_FILES=.
+lint: PYTHON_FILES=./infinity_emb
 lint_diff format_diff: PYTHON_FILES=$(shell git diff --relative=libs/infinity_emb --name-only --diff-filter=d master | grep -E '\.py$$|\.ipynb$$')
 
 lint lint_diff:
-	poetry run ruff .
-	[ "$(PYTHON_FILES)" = "" ] || poetry run black $(PYTHON_FILES) --check
+	poetry run ruff check .
 	[ "$(PYTHON_FILES)" = "" ] || poetry run mypy $(PYTHON_FILES)
 
 format format_diff:
-	[ "$(PYTHON_FILES)" = "" ] || poetry run black $(PYTHON_FILES)
-	[ "$(PYTHON_FILES)" = "" ] || poetry run ruff --select I --fix $(PYTHON_FILES)
+	[ "$(PYTHON_FILES)" = "" ] || poetry run ruff format $(PYTHON_FILES)
 
 template_docker:
 	jinja2 Dockerfile.jinja2 Docker.template.yaml --format=yaml -s amd > Dockerfile.amd_auto

libs/infinity_emb/infinity_emb/args.py

@@ -74,13 +74,9 @@ def __post_init__(self):
         if not isinstance(self.dtype, Dtype):
             object.__setattr__(self, "dtype", Dtype[self.dtype])
         if not isinstance(self.pooling_method, PoolingMethod):
-            object.__setattr__(
-                self, "pooling_method", PoolingMethod[self.pooling_method]
-            )
+            object.__setattr__(self, "pooling_method", PoolingMethod[self.pooling_method])
         if not isinstance(self.embedding_dtype, EmbeddingDtype):
-            object.__setattr__(
-                self, "embedding_dtype", EmbeddingDtype[self.embedding_dtype]
-            )
+            object.__setattr__(self, "embedding_dtype", EmbeddingDtype[self.embedding_dtype])
         if not self.served_model_name:
             object.__setattr__(
                 self,

libs/infinity_emb/infinity_emb/engine.py

@@ -52,9 +52,7 @@ def __init__(
 
         self.running = False
         self._running_sepamore: Optional[Semaphore] = None
-        self._model, self._min_inference_t, self._max_inference_t = select_model(
-            self._engine_args
-        )
+        self._model, self._min_inference_t, self._max_inference_t = select_model(self._engine_args)
 
     @classmethod
     def from_args(
@@ -130,9 +128,7 @@ def capabilities(self) -> set[ModelCapabilites]:
     def engine_args(self) -> EngineArgs:
         return self._engine_args
 
-    async def embed(
-        self, sentences: list[str]
-    ) -> tuple[list["EmbeddingReturnType"], int]:
+    async def embed(self, sentences: list[str]) -> tuple[list["EmbeddingReturnType"], int]:
         """embed multiple sentences
 
         Kwargs:
@@ -279,9 +275,7 @@ def __init__(self, engines: Iterable["AsyncEmbeddingEngine"]):
             set(engine.engine_args.served_model_name for engine in engines)
         ):
             raise ValueError("Engines must have unique model names")
-        self.engines_dict = {
-            engine.engine_args.served_model_name: engine for engine in engines
-        }
+        self.engines_dict = {engine.engine_args.served_model_name: engine for engine in engines}
 
     @classmethod
     def from_args(cls, engine_args_array: Iterable[EngineArgs]) -> "AsyncEngineArray":
@@ -358,9 +352,7 @@ async def rerank(
             list[float]: list of scores
             int: token usage
         """
-        return await self[model].rerank(
-            query=query, docs=docs, raw_scores=raw_scores, top_n=top_n
-        )
+        return await self[model].rerank(query=query, docs=docs, raw_scores=raw_scores, top_n=top_n)
 
     async def classify(
         self, *, model: str, sentences: list[str], raw_scores: bool = False

libs/infinity_emb/infinity_emb/env.py

@@ -61,9 +61,7 @@ def _optional_infinity_var(self, name: str, default: str = ""):
         self._debug(f"{name}=`{value}`")
         return value
 
-    def _optional_infinity_var_multiple(
-        self, name: str, default: list[str]
-    ) -> list[str]:
+    def _optional_infinity_var_multiple(self, name: str, default: list[str]) -> list[str]:
         self.__pre_fetch_env_manager()
         name = self.to_name(name)
         value = os.getenv(name)
@@ -133,9 +131,7 @@ def vector_disk_cache(self):
     @cached_property
     def lengths_via_tokenize(self):
         return self._to_bool_multiple(
-            self._optional_infinity_var_multiple(
-                "lengths_via_tokenize", default=["false"]
-            )
+            self._optional_infinity_var_multiple("lengths_via_tokenize", default=["false"])
         )
 
     @cached_property
@@ -152,9 +148,7 @@ def bettertransformer(self):
 
     @cached_property
     def preload_only(self):
-        return self._to_bool(
-            self._optional_infinity_var("preload_only", default="false")
-        )
+        return self._to_bool(self._optional_infinity_var("preload_only", default="false"))
 
     @cached_property
     def calibration_dataset_url(self):
@@ -198,9 +192,7 @@ def queue_size(self) -> int:
 
     @cached_property
     def permissive_cors(self):
-        return self._to_bool(
-            self._optional_infinity_var("permissive_cors", default="false")
-        )
+        return self._to_bool(self._optional_infinity_var("permissive_cors", default="false"))
 
     @cached_property
     def url_prefix(self):
@@ -223,19 +215,15 @@ def host(self):
     @cached_property
     def redirect_slash(self):
         route = self._optional_infinity_var("redirect_slash", default="/docs")
-        assert not route or route.startswith(
-            "/"
-        ), "INFINITY_REDIRECT_SLASH must start with /"
+        assert not route or route.startswith("/"), "INFINITY_REDIRECT_SLASH must start with /"
         return route
 
     @cached_property
     def log_level(self):
         return self._optional_infinity_var("log_level", default="info")
 
     def _typed_multiple(self, name: str, cls: type["EnumTypeLike"]) -> list["str"]:
-        result = self._optional_infinity_var_multiple(
-            name, default=[cls.default_value()]
-        )
+        result = self._optional_infinity_var_multiple(name, default=[cls.default_value()])
         assert all(cls(v) for v in result)
         return result
 

libs/infinity_emb/infinity_emb/fastapi_schemas/data_uri.py

@@ -192,9 +192,7 @@ def __get_pydantic_core_schema__(cls, source_type: Any, handler: Any) -> Any:
         from pydantic_core import core_schema
 
         # return core_schema.no_info_after_validator_function(cls, handler(str))
-        return core_schema.no_info_after_validator_function(
-            cls.validate, core_schema.str_schema()
-        )
+        return core_schema.no_info_after_validator_function(cls.validate, core_schema.str_schema())
 
     @classmethod
     def validate(
@@ -214,9 +212,7 @@ def validate(
         return m
 
     @classmethod
-    def __get_pydantic_json_schema__(
-        cls, schema: MutableMapping[str, Any], handler: Any
-    ) -> Any:
+    def __get_pydantic_json_schema__(cls, schema: MutableMapping[str, Any], handler: Any) -> Any:
         json_schema = handler(schema)
         json_schema.update(
             pattern=DATA_URI_REGEX,

libs/infinity_emb/infinity_emb/fastapi_schemas/pymodels.py

@@ -197,9 +197,10 @@ def to_embeddings_response(
                 raise ValueError(
                     f"model {engine_args.served_model_name} does not support base64 encoding, as it uses uint8-bitpacking with {engine_args.embedding_dtype}"
                 )
-            embeddings = [base64.b64encode(np.frombuffer(emb.astype(np.float32), dtype=np.float32)) for emb in embeddings]  # type: ignore
-        elif isinstance(embeddings, np.ndarray):
-            embeddings = embeddings.tolist()
+            embeddings = [
+                base64.b64encode(np.frombuffer(emb.astype(np.float32), dtype=np.float32))  # type: ignore
+                for emb in embeddings
+            ]  # type: ignore
         else:
             embeddings = [e.tolist() for e in embeddings]
         return dict(

libs/infinity_emb/infinity_emb/inference/batch_handler.py

@@ -121,9 +121,7 @@ def __init__(
                 " Consider increasing queue size"
             )
 
-    async def embed(
-        self, sentences: list[str]
-    ) -> tuple[list["EmbeddingReturnType"], int]:
+    async def embed(self, sentences: list[str]) -> tuple[list["EmbeddingReturnType"], int]:
         """Schedule a sentence to be embedded. Awaits until embedded.
 
         Args:
@@ -286,9 +284,7 @@ async def audio_embed(
         embeddings, usage = await self._schedule(items)
         return embeddings, usage
 
-    async def _schedule(
-        self, list_queueitem: Sequence[AbstractSingle]
-    ) -> tuple[list[Any], int]:
+    async def _schedule(self, list_queueitem: Sequence[AbstractSingle]) -> tuple[list[Any], int]:
         """adds list of items to the queue and awaits until these are completed."""
         prios, usage = await self._get_prios_usage(list_queueitem)
         new_prioqueue: list[PrioritizedQueueItem] = []
@@ -331,9 +327,7 @@ def overload_status(self) -> OverloadStatus:
             results_absolute=len(self._result_store),
         )
 
-    async def _get_prios_usage(
-        self, items: Sequence[AbstractSingle]
-    ) -> tuple[list[int], int]:
+    async def _get_prios_usage(self, items: Sequence[AbstractSingle]) -> tuple[list[int], int]:
         """get priorities and usage
 
         Args:
@@ -465,8 +459,7 @@ def _preprocess_batch(self):
                 # - or if many items are queued anyhow, so that a good batch
                 #   may be popped already.
                 if not self._feature_queue.empty() and (
-                    self._feature_queue.full()
-                    or (len(self._queue_prio) < self._max_batch_size * 4)
+                    self._feature_queue.full() or (len(self._queue_prio) < self._max_batch_size * 4)
                 ):
                     # add some stochastic delay
                     time.sleep(self._batch_delay)
@@ -550,8 +543,7 @@ def _postprocess_batch(self):
 
                 if (
                     self._postprocess_queue.empty()
-                    and self._last_inference
-                    < time.perf_counter() + self._batch_delay * 2
+                    and self._last_inference < time.perf_counter() + self._batch_delay * 2
                 ):
                     # 5 ms, assuming this is below
                     # 3-50ms for inference on avg.

libs/infinity_emb/infinity_emb/inference/caching_layer.py

@@ -4,6 +4,7 @@
 This file contains the experimental code for retrieving and storing result of embeddings to diskcache
 which may reduce latency.
 """
+
 import asyncio
 import queue
 import threading

libs/infinity_emb/infinity_emb/inference/select_model.py

@@ -22,9 +22,7 @@
 
 def get_engine_type_from_config(
     engine_args: EngineArgs,
-) -> Union[
-    EmbedderEngine, RerankEngine, PredictEngine, ImageEmbedEngine, AudioEmbedEngine
-]:
+) -> Union[EmbedderEngine, RerankEngine, PredictEngine, ImageEmbedEngine, AudioEmbedEngine]:
     """resolved the class of inference engine path from config.json of the repo."""
     if engine_args.engine in [InferenceEngine.debugengine]:
         return EmbedderEngine.from_inference_engine(engine_args.engine)
@@ -44,9 +42,7 @@ def get_engine_type_from_config(
     with open(config_path, "r") as f:
         config = json.load(f)
 
-    if any(
-        "SequenceClassification" in arch for arch in config.get("architectures", [])
-    ):
+    if any("SequenceClassification" in arch for arch in config.get("architectures", [])):
         id2label = config.get("id2label", {"0": "dummy"})
         if len(id2label) < 2:
             return RerankEngine.from_inference_engine(engine_args.engine)
@@ -81,9 +77,7 @@ def select_model(
         # size one, warm up warm start timings.
         loaded_engine.warmup(batch_size=engine_args.batch_size, n_tokens=1)
         # size one token
-        min_inference_t = min(
-            loaded_engine.warmup(batch_size=1, n_tokens=1)[1] for _ in range(10)
-        )
+        min_inference_t = min(loaded_engine.warmup(batch_size=1, n_tokens=1)[1] for _ in range(10))
         loaded_engine.warmup(batch_size=engine_args.batch_size, n_tokens=1)
         emb_per_sec_short, max_inference_t, log_msg = loaded_engine.warmup(
             batch_size=engine_args.batch_size, n_tokens=1

libs/infinity_emb/infinity_emb/infinity_server.py

@@ -112,10 +112,7 @@ async def kill_later(seconds: int):
                 await asyncio.sleep(seconds)
                 os.kill(os.getpid(), signal.SIGINT)
 
-            logger.info(
-                f"Preloaded configuration successfully. {engine_args_list} "
-                " -> exit ."
-            )
+            logger.info(f"Preloaded configuration successfully. {engine_args_list} " " -> exit .")
             asyncio.create_task(kill_later(3))
 
         yield
@@ -235,7 +232,7 @@ def _resolve_engine(model: str) -> "AsyncEmbeddingEngine":
         return engine
 
     def _resolve_mixed_input(
-        inputs: Union[DataURIorURL, list[DataURIorURL]]
+        inputs: Union[DataURIorURL, list[DataURIorURL]],
     ) -> list[Union[str, bytes]]:
         if hasattr(inputs, "host"):
             # if it is a single url
@@ -245,7 +242,8 @@ def _resolve_mixed_input(
         else:
             # is list, resolve to bytes or url
             urls_or_bytes: list[Union[str, bytes]] = [  # type: ignore
-                str(d) if hasattr(d, "host") else d.data for d in inputs  # type: ignore
+                str(d) if hasattr(d, "host") else d.data
+                for d in inputs  # type: ignore
             ]
         return urls_or_bytes
 
@@ -650,10 +648,7 @@ def typer_option_resolve(*args):
             if hasattr(args[0], "default") and hasattr(args[0], "envvar")
             else args[0]
         )
-    return (
-        a.default if (hasattr(a, "default") and hasattr(a, "envvar")) else a
-        for a in args
-    )
+    return (a.default if (hasattr(a, "default") and hasattr(a, "envvar")) else a for a in args)
 
 
 # CLI
@@ -694,9 +689,7 @@ def v1(
     ):
         """Infinity API ♾️  cli v1 - deprecated, consider use cli v2 via `infinity_emb v2`."""
         if api_key:
-            raise ValueError(
-                "api_key is not supported in `v1`. Please migrate to `v2`."
-            )
+            raise ValueError("api_key is not supported in `v1`. Please migrate to `v2`.")
         if not (
             embedding_dtype == EmbeddingDtype.float32
             or embedding_dtype == EmbeddingDtype.default_value()
@@ -820,12 +813,8 @@ def v2(
             **_construct("preload_only"),
             help="If true, only downloads models and verifies setup, then exit. Recommended for pre-caching the download in a Dockerfile.",
         ),
-        host: str = typer.Option(
-            **_construct("host"), help="host for the FastAPI uvicorn server"
-        ),
-        port: int = typer.Option(
-            **_construct("port"), help="port for the FastAPI uvicorn server"
-        ),
+        host: str = typer.Option(**_construct("host"), help="host for the FastAPI uvicorn server"),
+        port: int = typer.Option(**_construct("port"), help="port for the FastAPI uvicorn server"),
         url_prefix: str = typer.Option(
             **_construct("url_prefix"),
             callback=validate_url,
@@ -834,7 +823,9 @@ def v2(
         redirect_slash: str = typer.Option(
             **_construct("redirect_slash"), help="where to redirect `/` requests to."
         ),
-        log_level: UVICORN_LOG_LEVELS = typer.Option(**_construct("log_level"), help="console log level."),  # type: ignore
+        log_level: UVICORN_LOG_LEVELS = typer.Option(
+            **_construct("log_level"), help="console log level."
+        ),  # type: ignore
         permissive_cors: bool = typer.Option(
             **_construct("permissive_cors"), help="whether to allow permissive cors."
         ),