numpy._DTypeMeta object is not subscriptable When using the Chroma.from_documents #5424

pankai01 · 2023-05-30T04:25:23Z

pankai01
May 30, 2023

When I use the Chroma.from_documents,I encounter a type error:TypeError: 'numpy._DTypeMeta' object is not subscriptable
My code is here:

loader = DirectoryLoader('./data', glob='**/*.txt')
documents = loader.load()
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
split_docs = text_splitter.split_documents(documents)
embeddings = OpenAIEmbeddings()
docsearch = Chroma.from_documents(split_docs, embeddings, persist_directory="./chroma")

Python 3.9.16
numpy Version: 1.24.3
langchain Version: 0.0.184

and this the stack

TypeError Traceback (most recent call last)
Cell In[39], line 23
21 embeddings = OpenAIEmbeddings()
22 #将 document 通过 openai 的 embeddings 对象计算 embedding 向量信息并临时存入 chroma 向量数据库，用于后续匹配查询
---> 23 docsearch = Chroma.from_documents(split_docs, embeddings, persist_directory="./chroma")
25 #docsearch = Chroma(persist_directory="./chroma", embedding_function=embeddings)
26
27 #创建问答对象
(...)
31 #print(result)
32 #docsearch.persist()

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/langchain/vectorstores/chroma.py:433, in Chroma.from_documents(cls, documents, embedding, ids, collection_name, persist_directory, client_settings, client, **kwargs)
431 texts = [doc.page_content for doc in documents]
432 metadatas = [doc.metadata for doc in documents]
--> 433 return cls.from_texts(
434 texts=texts,
435 embedding=embedding,
436 metadatas=metadatas,
437 ids=ids,
438 collection_name=collection_name,
439 persist_directory=persist_directory,
440 client_settings=client_settings,
441 client=client,
442 )

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/langchain/vectorstores/chroma.py:394, in Chroma.from_texts(cls, texts, embedding, metadatas, ids, collection_name, persist_directory, client_settings, client, **kwargs)
364 @classmethod
365 def from_texts(
366 cls: Type[Chroma],
(...)
375 **kwargs: Any,
376 ) -> Chroma:
377 """Create a Chroma vectorstore from a raw documents.
378
379 If a persist_directory is specified, the collection will be persisted there.
(...)
392 Chroma: Chroma vectorstore.
393 """
--> 394 chroma_collection = cls(
395 collection_name=collection_name,
396 embedding_function=embedding,
397 persist_directory=persist_directory,
398 client_settings=client_settings,
399 client=client,
400 )
401 chroma_collection.add_texts(texts=texts, metadatas=metadatas, ids=ids)
402 return chroma_collection

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/langchain/vectorstores/chroma.py:69, in Chroma.init(self, collection_name, embedding_function, persist_directory, client_settings, collection_metadata, client)
67 """Initialize with Chroma client."""
68 try:
---> 69 import chromadb
70 import chromadb.config
71 except ImportError:

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/chromadb/init.py:1
----> 1 import chromadb.config
2 import logging
3 from chromadb.telemetry.events import ClientStartEvent

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/chromadb/config.py:6
4 import importlib
5 import logging
----> 6 import chromadb.db
7 import chromadb.api
8 import chromadb.telemetry

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/chromadb/db/init.py:4
2 from typing import List, Sequence, Optional, Tuple
3 from uuid import UUID
----> 4 import numpy.typing as npt
5 from chromadb.api.types import (
6 Embeddings,
7 Documents,
(...)
12 WhereDocument,
13 )
16 class DB(ABC):

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/numpy/typing/init.py:158
1 """
2 ============================
3 Typing (:mod:numpy.typing)
(...)
153
154 """
155 # NOTE: The API section will be appended with additional entries
156 # further down in this file
--> 158 from numpy._typing import (
159 ArrayLike,
160 DTypeLike,
161 NBitBase,
162 NDArray,
163 )
165 all = ["ArrayLike", "DTypeLike", "NBitBase", "NDArray"]
167 if doc is not None:

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/numpy/_typing/init.py:164
148 from ._scalars import (
149 _CharLike_co as _CharLike_co,
150 _BoolLike_co as _BoolLike_co,
(...)
158 _VoidLike_co as _VoidLike_co,
159 )
160 from ._shape import (
161 _Shape as _Shape,
162 _ShapeLike as _ShapeLike,
163 )
--> 164 from ._dtype_like import (
165 DTypeLike as DTypeLike,
166 _DTypeLike as _DTypeLike,
167 _SupportsDType as _SupportsDType,
168 _VoidDTypeLike as _VoidDTypeLike,
169 _DTypeLikeBool as _DTypeLikeBool,
170 _DTypeLikeUInt as _DTypeLikeUInt,
171 _DTypeLikeInt as _DTypeLikeInt,
172 _DTypeLikeFloat as _DTypeLikeFloat,
173 _DTypeLikeComplex as _DTypeLikeComplex,
174 _DTypeLikeTD64 as _DTypeLikeTD64,
175 _DTypeLikeDT64 as _DTypeLikeDT64,
176 _DTypeLikeObject as _DTypeLikeObject,
177 _DTypeLikeVoid as _DTypeLikeVoid,
178 _DTypeLikeStr as _DTypeLikeStr,
179 _DTypeLikeBytes as _DTypeLikeBytes,
180 _DTypeLikeComplex_co as _DTypeLikeComplex_co,
181 )
182 from ._array_like import (
183 ArrayLike as ArrayLike,
184 _ArrayLike as _ArrayLike,
(...)
202 _UnknownType as _UnknownType,
203 )
204 from ._generic_alias import (
205 NDArray as NDArray,
206 _DType as _DType,
207 _GenericAlias as _GenericAlias,
208 )

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/numpy/_typing/_dtype_like.py:17
14 import numpy as np
16 from ._shape import _ShapeLike
---> 17 from ._generic_alias import _DType as DType
19 from ._char_codes import (
20 _BoolCodes,
21 _UInt8Codes,
(...)
58 _ObjectCodes,
59 )
61 _SCT = TypeVar("_SCT", bound=np.generic)

File ~/miniconda3/envs/d2l/lib/python3.9/site-packages/numpy/_typing/_generic_alias.py:241
238 ScalarType = TypeVar("ScalarType", bound=np.generic, covariant=True)
240 if TYPE_CHECKING or sys.version_info >= (3, 9):
--> 241 _DType = np.dtype[ScalarType]
242 NDArray = np.ndarray[Any, np.dtype[ScalarType]]
243 else:

TypeError: 'numpy._DTypeMeta' object is not subscriptable