huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-09 19:34:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

black

Browse Source
This commit is contained in:
Daniël de Kok 2025-08-28 14:14:35 +00:00
parent db187caf6b
commit 8834aaf46d
10 changed files with 22 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
integration-tests/models/test_grammar_response_format_llama.py
View File

@ -43,7 +43,10 @@ async def test_grammar_response_format_llama_json(llama_grammar, response_snapsh
], ],
"seed": 42, "seed": 42,
"max_tokens": 500, "max_tokens": 500,
"response_format": {"type": "json_object", "value": Weather.model_json_schema()}, "response_format": {
"type": "json_object",
"value": Weather.model_json_schema(),
},
} }
# send the request # send the request
response = requests.post( response = requests.post(
@ -75,7 +78,11 @@ async def test_grammar_response_format_llama_json(llama_grammar, response_snapsh
json_payload["response_format"] = { json_payload["response_format"] = {
"type": "json_schema", "type": "json_schema",
"value": {"name": "weather", "strict": True, "schema": Weather.model_json_schema()}, "value": {
"name": "weather",
"strict": True,
"schema": Weather.model_json_schema(),
},
} }
response = requests.post( response = requests.post(
f"{llama_grammar.base_url}/v1/chat/completions", f"{llama_grammar.base_url}/v1/chat/completions",
@ -119,7 +126,10 @@ async def test_grammar_response_format_llama_error_if_tools_not_installed(
"seed": 42, "seed": 42,
"max_tokens": 500, "max_tokens": 500,
"tools": [], "tools": [],
"response_format": {"type": "json_object", "value": Weather.model_json_schema()}, "response_format": {
"type": "json_object",
"value": Weather.model_json_schema(),
},
}, },
) )

2
server/text_generation_server/utils/segments.py
View File

@ -9,7 +9,7 @@ import numpy as np
def find_segments( def find_segments(
adapter_indices: Union[torch.Tensor, List[int]] adapter_indices: Union[torch.Tensor, List[int]],
) -> Tuple[List[int], List[int]]: ) -> Tuple[List[int], List[int]]:
if isinstance(adapter_indices, torch.Tensor): if isinstance(adapter_indices, torch.Tensor):
adapter_indices = adapter_indices.cpu().numpy() adapter_indices = adapter_indices.cpu().numpy()