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

[Core][VLM] Add precise multi-modal placeholder tracking (#8346)

Browse Source 
Signed-off-by: Peter Salas <peter@fixie.ai>
This commit is contained in:
Peter Salas
2024-11-01 16:21:10 -07:00
committed by GitHub
parent d151fde834
commit 6c0b7f548d
53 changed files with 913 additions and 281 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/worker/test_model_input.py
View File

@@ -73,6 +73,7 @@ def test_model_runner_input():
num_prefill_tokens=2,
num_decode_tokens=3,
slot_mapping=torch.zeros(1),
multi_modal_placeholder_index_maps=None,
)
model_input = ModelInputForGPUWithSamplingMetadata(
input_tokens=torch.ones(10),
@@ -124,6 +125,7 @@ def test_embedding_model_runner_input():
num_prefill_tokens=2,
num_decode_tokens=3,
slot_mapping=torch.zeros(1),
multi_modal_placeholder_index_maps=None,
)
model_input = ModelInputForGPUWithPoolingMetadata(
input_tokens=torch.ones(10),
@@ -174,6 +176,7 @@ def test_multi_step_model_runner_input():
num_prefill_tokens=2,
num_decode_tokens=3,
slot_mapping=torch.zeros(1),
multi_modal_placeholder_index_maps=None,
)
frozen_model_input = ModelInputForGPUWithSamplingMetadata(
input_tokens=torch.ones(10),