EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[CI][UT]enable multimodal ut (#158)

Browse Source 
enable multimodal ut

---------

Signed-off-by: zouyida <zouyida@huawei.com>
This commit is contained in:
zouyida2002
2025-02-27 14:14:43 +08:00
committed by GitHub
parent 94483775e1
commit 94cd66bba7
2 changed files with 6 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
vllm_ascend/model_runner.py
View File

@@ -695,6 +695,7 @@ class ModelInputForNPUBuilder(ModelRunnerInputBuilderBase[ModelInputForNPU]):
assert image_grid_thw is not None or video_grid_thw is not None, (
"mrope embedding type requires multi-modal input mapper "
"returns 'image_grid_thw' or 'video_grid_thw'.")
second_per_grid_ts = mm_kwargs.get("second_per_grid_ts", None)
hf_config = self.runner.model_config.hf_config
@@ -707,14 +708,10 @@ class ModelInputForNPUBuilder(ModelRunnerInputBuilderBase[ModelInputForNPU]):
mrope_input_positions, mrope_position_delta = \
MRotaryEmbedding.get_input_positions(
token_ids,
hf_config,
image_grid_thw=image_grid_thw,
video_grid_thw=video_grid_thw,
image_token_id=hf_config.image_token_id,
video_token_id=hf_config.video_token_id,
vision_start_token_id=hf_config.vision_start_token_id,
vision_end_token_id=hf_config.vision_end_token_id,
spatial_merge_size=hf_config.vision_config.
spatial_merge_size,
second_per_grid_ts=second_per_grid_ts,
context_len=inter_data.context_lens[seq_idx],
seq_len=inter_data.seq_lens[seq_idx],
)