huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 12:24:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: refactor tests to support completions snapshot

Browse Source
This commit is contained in:
drbh 2024-04-16 15:53:27 +00:00
parent 0b82080849
commit ef2363cd3a
4 changed files with 647 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
clients/python/text_generation/types.py
View File

@ -59,6 +59,17 @@ class ChatCompletionComplete(BaseModel):
usage: Optional[Any] = None usage: Optional[Any] = None
class CompletionComplete(BaseModel):
# Index of the chat completion
index: int
# Message associated with the chat completion
text: str
# Log probabilities for the chat completion
logprobs: Optional[Any]
# Reason for completion
finish_reason: str
class Function(BaseModel): class Function(BaseModel):
name: Optional[str] name: Optional[str]
arguments: str arguments: str
@ -104,6 +115,16 @@ class ChatComplete(BaseModel):
usage: Any usage: Any
class CompletionComplete(BaseModel):
# Completion details
id: str
object: str
created: int
model: str
system_fingerprint: str
choices: List[CompletionComplete]
class ChatRequest(BaseModel): class ChatRequest(BaseModel):
# Model identifier # Model identifier
model: str model: str

22
integration-tests/conftest.py
View File

@ -9,6 +9,7 @@ import json
import math import math
import time import time
import random import random
import re
from docker.errors import NotFound from docker.errors import NotFound
from typing import Optional, List, Dict from typing import Optional, List, Dict
@ -26,6 +27,7 @@ from text_generation.types import (
ChatComplete, ChatComplete,
ChatCompletionChunk, ChatCompletionChunk,
ChatCompletionComplete, ChatCompletionComplete,
CompletionComplete,
) )
DOCKER_IMAGE = os.getenv("DOCKER_IMAGE", None) DOCKER_IMAGE = os.getenv("DOCKER_IMAGE", None)
@ -69,12 +71,12 @@ class ResponseComparator(JSONSnapshotExtension):
data = json.loads(data) data = json.loads(data)
if isinstance(data, Dict) and "choices" in data: if isinstance(data, Dict) and "choices" in data:
choices = data["choices"] choices = data["choices"]
if ( print(choices)
isinstance(choices, List) if isinstance(choices, List) and len(choices) >= 1:
and len(choices) >= 1 if "delta" in choices[0]:
and "delta" in choices[0] return ChatCompletionChunk(**data)
): if "text" in choices[0]:
return ChatCompletionChunk(**data) return CompletionComplete(**data)
return ChatComplete(**data) return ChatComplete(**data)
if isinstance(data, Dict): if isinstance(data, Dict):
@ -161,6 +163,9 @@ class ResponseComparator(JSONSnapshotExtension):
) )
) )
def eq_completion(response: ChatComplete, other: ChatComplete) -> bool:
return response.choices[0].text == other.choices[0].text
def eq_chat_complete(response: ChatComplete, other: ChatComplete) -> bool: def eq_chat_complete(response: ChatComplete, other: ChatComplete) -> bool:
return ( return (
response.choices[0].message.content == other.choices[0].message.content response.choices[0].message.content == other.choices[0].message.content
@ -184,6 +189,11 @@ class ResponseComparator(JSONSnapshotExtension):
if not isinstance(snapshot_data, List): if not isinstance(snapshot_data, List):
snapshot_data = [snapshot_data] snapshot_data = [snapshot_data]
if isinstance(serialized_data[0], CompletionComplete):
return len(snapshot_data) == len(serialized_data) and all(
[eq_completion(r, o) for r, o in zip(serialized_data, snapshot_data)]
)
if isinstance(serialized_data[0], ChatComplete): if isinstance(serialized_data[0], ChatComplete):
return len(snapshot_data) == len(serialized_data) and all( return len(snapshot_data) == len(serialized_data) and all(
[eq_chat_complete(r, o) for r, o in zip(serialized_data, snapshot_data)] [eq_chat_complete(r, o) for r, o in zip(serialized_data, snapshot_data)]

603
integration-tests/models/__snapshots__/test_completion_prompts/test_flash_llama_completion_many_prompts_stream.json
View File

@ -1 +1,602 @@
"<ClientResponse(http://localhost:8041/v1/completions) [200 OK]>\n<CIMultiDictProxy('Content-Type': 'text/event-stream', 'Cache-Control': 'no-cache', 'x-compute-type': '1-nvidia-a10g', 'x-compute-characters': '72', 'x-accel-buffering': 'no', 'Access-Control-Allow-Origin': '*', 'Vary': 'origin', 'Vary': 'access-control-request-method', 'Vary': 'access-control-request-headers', 'Transfer-Encoding': 'chunked', 'Date': 'Thu, 11 Apr 2024 22:43:33 GMT')>\n" [
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": "\n"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": "\n"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": "\n"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "hd"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": "2"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": "2"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": "2"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "2"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": "0"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": "."
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": "."
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "\n"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": "."
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": " And"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": " And"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "def"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": " What"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": " away"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": " what"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": " mai"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": " is"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": " we"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": " is"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "hd"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": " the"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": " go"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": " the"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "\n"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": " temperature"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": ","
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": " capital"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "\t"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": " in"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": " sc"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": " of"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": "p"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 0,
"logprobs": null,
"text": " New"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 1,
"logprobs": null,
"text": "rap"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 2,
"logprobs": null,
"text": " Ukraine"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
},
{
"choices": [
{
"finish_reason": "",
"index": 3,
"logprobs": null,
"text": " text"
}
],
"created": 1713282708,
"id": "",
"model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
"object": "text_completion",
"system_fingerprint": "1.4.5-native"
}
]

9
integration-tests/models/test_completion_prompts.py
View File

@ -3,6 +3,10 @@ import requests
import json import json
from aiohttp import ClientSession from aiohttp import ClientSession
from text_generation.types import (
CompletionComplete,
)
@pytest.fixture(scope="module") @pytest.fixture(scope="module")
def flash_llama_completion_handle(launcher): def flash_llama_completion_handle(launcher):
@ -82,6 +86,7 @@ async def test_flash_llama_completion_many_prompts_stream(
url = f"{flash_llama_completion.base_url}/v1/completions" url = f"{flash_llama_completion.base_url}/v1/completions"
chunks = []
async with ClientSession(headers=flash_llama_completion.headers) as session: async with ClientSession(headers=flash_llama_completion.headers) as session:
async with session.post(url, json=request) as response: async with session.post(url, json=request) as response:
# iterate over the stream # iterate over the stream
@ -96,8 +101,10 @@ async def test_flash_llama_completion_many_prompts_stream(
chunk = [json.loads(c) for c in chunk] chunk = [json.loads(c) for c in chunk]
for c in chunk: for c in chunk:
print(c)
chunks.append(CompletionComplete(**c))
assert "choices" in c assert "choices" in c
assert 0 <= c["choices"][0]["index"] <= 4 assert 0 <= c["choices"][0]["index"] <= 4
assert response.status == 200 assert response.status == 200
assert response == response_snapshot assert chunks == response_snapshot