EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Support decode token logprobs (#130)

Browse Source
This commit is contained in:
Cody Yu
2024-02-06 12:24:55 -08:00
committed by GitHub
parent ee1df26a77
commit a7334aeea1
10 changed files with 233 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
test/srt/test_openai_server.py
View File

@@ -18,15 +18,26 @@ import argparse
import openai
def test_completion(args):
def test_completion(args, echo, logprobs):
client = openai.Client(api_key="EMPTY", base_url=args.base_url)
response = client.completions.create(
model="default",
prompt="The capital of France is",
temperature=0,
max_tokens=32,
echo=echo,
logprobs=logprobs,
)
text = response.choices[0].text
print(response.choices[0].text)
if echo:
assert text.startswith("The capital of France is")
if logprobs:
assert response.choices[0].logprobs
if echo:
assert response.choices[0].logprobs.token_logprobs[0] == None
else:
assert response.choices[0].logprobs.token_logprobs[0] != None
assert response.id
assert response.created
assert response.usage.prompt_tokens > 0
@@ -34,7 +45,7 @@ def test_completion(args):
assert response.usage.total_tokens > 0
def test_completion_stream(args):
def test_completion_stream(args, echo, logprobs):
client = openai.Client(api_key="EMPTY", base_url=args.base_url)
response = client.completions.create(
model="default",
@@ -42,9 +53,23 @@ def test_completion_stream(args):
temperature=0,
max_tokens=32,
stream=True,
echo=echo,
logprobs=logprobs,
)
first = True
for r in response:
print(r.choices[0].text, end="", flush=True)
if first:
if echo:
assert r.choices[0].text.startswith("The capital of France is")
first = False
if logprobs:
print(
f"{r.choices[0].text:12s}\t"
f"{r.choices[0].logprobs.token_logprobs}",
flush=True
)
else:
print(r.choices[0].text, end="", flush=True)
assert r.id
assert r.usage.prompt_tokens > 0
assert r.usage.completion_tokens > 0
@@ -135,8 +160,14 @@ if __name__ == "__main__":
)
args = parser.parse_args()
test_completion(args)
test_completion_stream(args)
test_completion(args, echo=False, logprobs=False)
test_completion(args, echo=True, logprobs=False)
test_completion(args, echo=False, logprobs=True)
test_completion(args, echo=True, logprobs=True)
test_completion_stream(args, echo=False, logprobs=False)
test_completion_stream(args, echo=True, logprobs=False)
test_completion_stream(args, echo=False, logprobs=True)
test_completion_stream(args, echo=True, logprobs=True)
test_chat_completion(args)
test_chat_completion_stream(args)
if args.test_image: