Spaces:
Build error
Build error
File size: 7,611 Bytes
5a29263 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 |
import base64
import struct
import pytest
from openai import OpenAI
from utils import *
server = ServerPreset.bert_bge_small()
EPSILON = 1e-3
@pytest.fixture(scope="module", autouse=True)
def create_server():
global server
server = ServerPreset.bert_bge_small()
def test_embedding_single():
global server
server.pooling = 'last'
server.start()
res = server.make_request("POST", "/v1/embeddings", data={
"input": "I believe the meaning of life is",
})
assert res.status_code == 200
assert len(res.body['data']) == 1
assert 'embedding' in res.body['data'][0]
assert len(res.body['data'][0]['embedding']) > 1
# make sure embedding vector is normalized
assert abs(sum([x ** 2 for x in res.body['data'][0]['embedding']]) - 1) < EPSILON
def test_embedding_multiple():
global server
server.pooling = 'last'
server.start()
res = server.make_request("POST", "/v1/embeddings", data={
"input": [
"I believe the meaning of life is",
"Write a joke about AI from a very long prompt which will not be truncated",
"This is a test",
"This is another test",
],
})
assert res.status_code == 200
assert len(res.body['data']) == 4
for d in res.body['data']:
assert 'embedding' in d
assert len(d['embedding']) > 1
@pytest.mark.parametrize(
"input,is_multi_prompt",
[
# do not crash on empty input
("", False),
# single prompt
("string", False),
([12, 34, 56], False),
([12, 34, "string", 56, 78], False),
# multiple prompts
(["string1", "string2"], True),
(["string1", [12, 34, 56]], True),
([[12, 34, 56], [12, 34, 56]], True),
([[12, 34, 56], [12, "string", 34, 56]], True),
]
)
def test_embedding_mixed_input(input, is_multi_prompt: bool):
global server
server.start()
res = server.make_request("POST", "/v1/embeddings", data={"input": input})
assert res.status_code == 200
data = res.body['data']
if is_multi_prompt:
assert len(data) == len(input)
for d in data:
assert 'embedding' in d
assert len(d['embedding']) > 1
else:
assert 'embedding' in data[0]
assert len(data[0]['embedding']) > 1
def test_embedding_pooling_none():
global server
server.pooling = 'none'
server.start()
res = server.make_request("POST", "/embeddings", data={
"input": "hello hello hello",
})
assert res.status_code == 200
assert 'embedding' in res.body[0]
assert len(res.body[0]['embedding']) == 5 # 3 text tokens + 2 special
# make sure embedding vector is not normalized
for x in res.body[0]['embedding']:
assert abs(sum([x ** 2 for x in x]) - 1) > EPSILON
def test_embedding_pooling_none_oai():
global server
server.pooling = 'none'
server.start()
res = server.make_request("POST", "/v1/embeddings", data={
"input": "hello hello hello",
})
# /v1/embeddings does not support pooling type 'none'
assert res.status_code == 400
assert "error" in res.body
def test_embedding_openai_library_single():
global server
server.pooling = 'last'
server.start()
client = OpenAI(api_key="dummy", base_url=f"http://{server.server_host}:{server.server_port}/v1")
res = client.embeddings.create(model="text-embedding-3-small", input="I believe the meaning of life is")
assert len(res.data) == 1
assert len(res.data[0].embedding) > 1
def test_embedding_openai_library_multiple():
global server
server.pooling = 'last'
server.start()
client = OpenAI(api_key="dummy", base_url=f"http://{server.server_host}:{server.server_port}/v1")
res = client.embeddings.create(model="text-embedding-3-small", input=[
"I believe the meaning of life is",
"Write a joke about AI from a very long prompt which will not be truncated",
"This is a test",
"This is another test",
])
assert len(res.data) == 4
for d in res.data:
assert len(d.embedding) > 1
def test_embedding_error_prompt_too_long():
global server
server.pooling = 'last'
server.start()
res = server.make_request("POST", "/v1/embeddings", data={
"input": "This is a test " * 512,
})
assert res.status_code != 200
assert "too large" in res.body["error"]["message"]
def test_same_prompt_give_same_result():
server.pooling = 'last'
server.start()
res = server.make_request("POST", "/v1/embeddings", data={
"input": [
"I believe the meaning of life is",
"I believe the meaning of life is",
"I believe the meaning of life is",
"I believe the meaning of life is",
"I believe the meaning of life is",
],
})
assert res.status_code == 200
assert len(res.body['data']) == 5
for i in range(1, len(res.body['data'])):
v0 = res.body['data'][0]['embedding']
vi = res.body['data'][i]['embedding']
for x, y in zip(v0, vi):
assert abs(x - y) < EPSILON
@pytest.mark.parametrize(
"content,n_tokens",
[
("I believe the meaning of life is", 9),
("This is a test", 6),
]
)
def test_embedding_usage_single(content, n_tokens):
global server
server.start()
res = server.make_request("POST", "/v1/embeddings", data={"input": content})
assert res.status_code == 200
assert res.body['usage']['prompt_tokens'] == res.body['usage']['total_tokens']
assert res.body['usage']['prompt_tokens'] == n_tokens
def test_embedding_usage_multiple():
global server
server.start()
res = server.make_request("POST", "/v1/embeddings", data={
"input": [
"I believe the meaning of life is",
"I believe the meaning of life is",
],
})
assert res.status_code == 200
assert res.body['usage']['prompt_tokens'] == res.body['usage']['total_tokens']
assert res.body['usage']['prompt_tokens'] == 2 * 9
def test_embedding_openai_library_base64():
server.start()
test_input = "Test base64 embedding output"
# get embedding in default format
res = server.make_request("POST", "/v1/embeddings", data={
"input": test_input
})
assert res.status_code == 200
vec0 = res.body["data"][0]["embedding"]
# get embedding in base64 format
res = server.make_request("POST", "/v1/embeddings", data={
"input": test_input,
"encoding_format": "base64"
})
assert res.status_code == 200
assert "data" in res.body
assert len(res.body["data"]) == 1
embedding_data = res.body["data"][0]
assert "embedding" in embedding_data
assert isinstance(embedding_data["embedding"], str)
# Verify embedding is valid base64
decoded = base64.b64decode(embedding_data["embedding"])
# Verify decoded data can be converted back to float array
float_count = len(decoded) // 4 # 4 bytes per float
floats = struct.unpack(f'{float_count}f', decoded)
assert len(floats) > 0
assert all(isinstance(x, float) for x in floats)
assert len(floats) == len(vec0)
# make sure the decoded data is the same as the original
for x, y in zip(floats, vec0):
assert abs(x - y) < EPSILON
|