[Core] Use individual MM items in P0/P1 cache and model runner (#22570)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-08-13 22:18:07 +08:00
committed by GitHub
parent 20d65aa755
commit 19b927e52d
24 changed files with 549 additions and 486 deletions

View File

@@ -53,7 +53,7 @@ def test_incremental_detokenization(request_output_kind: RequestOutputKind,
EngineCoreRequest(request_id=f"request-{idx}",
prompt_token_ids=prompt_tokens,
arrival_time=0,
mm_inputs=None,
mm_kwargs=None,
mm_hashes=None,
mm_placeholders=None,
eos_token_id=None,
@@ -402,7 +402,7 @@ def test_logprobs_processor(request_output_kind: RequestOutputKind,
EngineCoreRequest(request_id=request_id_list[idx],
prompt_token_ids=prompt_tokens,
arrival_time=0,
mm_inputs=None,
mm_kwargs=None,
mm_hashes=None,
mm_placeholders=None,
eos_token_id=None,
@@ -567,7 +567,7 @@ def test_stop_token(include_stop_str_in_output: bool,
request_id=request_id,
prompt_token_ids=prompt_tokens,
arrival_time=0,
mm_inputs=None,
mm_kwargs=None,
mm_hashes=None,
mm_placeholders=None,
eos_token_id=eos_token_id,
@@ -666,7 +666,7 @@ def test_stop_string(include_stop_str_in_output: bool,
request_id=request_id_list[idx],
prompt_token_ids=prompt_tokens,
arrival_time=0,
mm_inputs=None,
mm_kwargs=None,
mm_hashes=None,
mm_placeholders=None,
eos_token_id=None,
@@ -782,7 +782,7 @@ def test_iteration_stats(dummy_test_vectors):
request_id=f"request-{idx}",
prompt_token_ids=prompt_tokens,
arrival_time=0,
mm_inputs=None,
mm_kwargs=None,
mm_hashes=None,
mm_placeholders=None,
eos_token_id=None,