mirror of
https://github.com/hwchase17/langchain.git
synced 2025-08-17 00:17:47 +00:00
56499cf58b
From what I can tell response using SDK is not deterministic:
```python
import numpy as np
import openai
documents = ["disallowed special token '<|endoftext|>'"]
model = "text-embedding-ada-002"
direct_output_1 = (
openai.OpenAI()
.embeddings.create(input=documents, model=model)
.data[0]
.embedding
)
for i in range(10):
direct_output_2 = (
openai.OpenAI()
.embeddings.create(input=documents, model=model)
.data[0]
.embedding
)
print(f"{i}: {np.isclose(direct_output_1, direct_output_2).all()}")
```
```
0: True
1: True
2: True
3: True
4: False
5: True
6: True
7: True
8: True
9: True
```
See related discussion here:
https://community.openai.com/t/can-text-embedding-ada-002-be-made-deterministic/318054
Found the same result using `"text-embedding-3-small"`.
134 lines
4.1 KiB
Python
134 lines
4.1 KiB
Python
"""Test azure openai embeddings."""
|
|
|
|
import os
|
|
from typing import Any
|
|
|
|
import numpy as np
|
|
import openai
|
|
import pytest
|
|
|
|
from langchain_openai import AzureOpenAIEmbeddings
|
|
|
|
OPENAI_API_VERSION = os.environ.get("AZURE_OPENAI_API_VERSION", "")
|
|
OPENAI_API_BASE = os.environ.get("AZURE_OPENAI_API_BASE", "")
|
|
OPENAI_API_KEY = os.environ.get("AZURE_OPENAI_API_KEY", "")
|
|
DEPLOYMENT_NAME = os.environ.get(
|
|
"AZURE_OPENAI_DEPLOYMENT_NAME",
|
|
os.environ.get("AZURE_OPENAI_EMBEDDINGS_DEPLOYMENT_NAME", ""),
|
|
)
|
|
print
|
|
|
|
|
|
def _get_embeddings(**kwargs: Any) -> AzureOpenAIEmbeddings:
|
|
return AzureOpenAIEmbeddings( # type: ignore[call-arg]
|
|
azure_deployment=DEPLOYMENT_NAME,
|
|
api_version=OPENAI_API_VERSION,
|
|
azure_endpoint=OPENAI_API_BASE,
|
|
openai_api_key=OPENAI_API_KEY,
|
|
**kwargs,
|
|
)
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
def test_azure_openai_embedding_documents() -> None:
|
|
"""Test openai embeddings."""
|
|
documents = ["foo bar"]
|
|
embedding = _get_embeddings()
|
|
output = embedding.embed_documents(documents)
|
|
assert len(output) == 1
|
|
assert len(output[0]) == 1536
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
def test_azure_openai_embedding_documents_multiple() -> None:
|
|
"""Test openai embeddings."""
|
|
documents = ["foo bar", "bar foo", "foo"]
|
|
embedding = _get_embeddings(chunk_size=2)
|
|
embedding.embedding_ctx_length = 8191
|
|
output = embedding.embed_documents(documents)
|
|
assert embedding.chunk_size == 2
|
|
assert len(output) == 3
|
|
assert len(output[0]) == 1536
|
|
assert len(output[1]) == 1536
|
|
assert len(output[2]) == 1536
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
def test_azure_openai_embedding_documents_chunk_size() -> None:
|
|
"""Test openai embeddings."""
|
|
documents = ["foo bar"] * 20
|
|
embedding = _get_embeddings()
|
|
embedding.embedding_ctx_length = 8191
|
|
output = embedding.embed_documents(documents)
|
|
# Max 2048 chunks per batch on Azure OpenAI embeddings
|
|
assert embedding.chunk_size == 2048
|
|
assert len(output) == 20
|
|
assert all([len(out) == 1536 for out in output])
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
async def test_azure_openai_embedding_documents_async_multiple() -> None:
|
|
"""Test openai embeddings."""
|
|
documents = ["foo bar", "bar foo", "foo"]
|
|
embedding = _get_embeddings(chunk_size=2)
|
|
embedding.embedding_ctx_length = 8191
|
|
output = await embedding.aembed_documents(documents)
|
|
assert len(output) == 3
|
|
assert len(output[0]) == 1536
|
|
assert len(output[1]) == 1536
|
|
assert len(output[2]) == 1536
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
def test_azure_openai_embedding_query() -> None:
|
|
"""Test openai embeddings."""
|
|
document = "foo bar"
|
|
embedding = _get_embeddings()
|
|
output = embedding.embed_query(document)
|
|
assert len(output) == 1536
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
async def test_azure_openai_embedding_async_query() -> None:
|
|
"""Test openai embeddings."""
|
|
document = "foo bar"
|
|
embedding = _get_embeddings()
|
|
output = await embedding.aembed_query(document)
|
|
assert len(output) == 1536
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
def test_azure_openai_embedding_with_empty_string() -> None:
|
|
"""Test openai embeddings with empty string."""
|
|
|
|
document = ["", "abc"]
|
|
embedding = _get_embeddings()
|
|
output = embedding.embed_documents(document)
|
|
assert len(output) == 2
|
|
assert len(output[0]) == 1536
|
|
expected_output = (
|
|
openai.AzureOpenAI(
|
|
api_version=OPENAI_API_VERSION,
|
|
api_key=OPENAI_API_KEY,
|
|
azure_endpoint=OPENAI_API_BASE,
|
|
azure_deployment=DEPLOYMENT_NAME,
|
|
) # type: ignore
|
|
.embeddings.create(input="", model="text-embedding-ada-002")
|
|
.data[0]
|
|
.embedding
|
|
)
|
|
assert np.allclose(output[0], expected_output, atol=0.001)
|
|
assert len(output[1]) == 1536
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
def test_embed_documents_normalized() -> None:
|
|
output = _get_embeddings().embed_documents(["foo walked to the market"])
|
|
assert np.isclose(np.linalg.norm(output[0]), 1.0)
|
|
|
|
|
|
@pytest.mark.scheduled
|
|
def test_embed_query_normalized() -> None:
|
|
output = _get_embeddings().embed_query("foo walked to the market")
|
|
assert np.isclose(np.linalg.norm(output), 1.0)
|