biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Core] Prevent side-channel attacks via cache salting (#17045)

Browse Source 
Signed-off-by: Marko Rosenmueller <5467316+dr75@users.noreply.github.com>
This commit is contained in:
Marko Rosenmueller
2025-04-30 14:27:21 +02:00
committed by GitHub
parent a7d5b016bd
commit 77073c77bc
18 changed files with 328 additions and 126 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
tests/v1/engine/test_output_processor.py
View File

@@ -57,6 +57,7 @@ def test_incremental_detokenization(request_output_kind: RequestOutputKind,
mm_placeholders=None,
eos_token_id=None,
lora_request=None,
cache_salt=None,
sampling_params=SamplingParams(
skip_special_tokens=False,
spaces_between_special_tokens=False,
@@ -403,6 +404,7 @@ def test_logprobs_processor(request_output_kind: RequestOutputKind,
mm_placeholders=None,
eos_token_id=None,
lora_request=None,
cache_salt=None,
sampling_params=SamplingParams(
skip_special_tokens=False,
spaces_between_special_tokens=False,
@@ -503,7 +505,7 @@ def test_stop_token(include_stop_str_in_output: bool,
reason should be "stop" (i.e. first control token causes stop
and is represented in output text)
* else, the detokenized string should be
* else, the detokenized string should be
<token><token>...<token> and the finish reason should be "stop"
(i.e. first control token causes stop but is not represented
in output text.)
@@ -565,6 +567,7 @@ def test_stop_token(include_stop_str_in_output: bool,
mm_placeholders=None,
eos_token_id=eos_token_id,
lora_request=None,
cache_salt=None,
sampling_params=SamplingParams(
skip_special_tokens=False,
spaces_between_special_tokens=False,
@@ -661,6 +664,7 @@ def test_stop_string(include_stop_str_in_output: bool,
mm_placeholders=None,
eos_token_id=None,
lora_request=None,
cache_salt=None,
sampling_params=SamplingParams(
skip_special_tokens=False,
spaces_between_special_tokens=False,
@@ -774,6 +778,7 @@ def test_iteration_stats(dummy_test_vectors):
mm_placeholders=None,
eos_token_id=None,
lora_request=None,
cache_salt=None,
sampling_params=SamplingParams(),
) for idx, prompt_tokens in enumerate(dummy_test_vectors.prompt_tokens)
]