Skip to content

[ModelRunner][MultiModal] Remove legacy input mapper/processor from V0 #951

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jun 3, 2025
Merged

[ModelRunner][MultiModal] Remove legacy input mapper/processor from V0 #951

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
db13d73
update multi-modal code in V0 ModelRunner
shen-shanshan May 26, 2025
3b76adb
Merge branch 'main' into mm
wangxiyuan Jun 3, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 5 additions & 17 deletions vllm_ascend/worker/model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -391,7 +391,6 @@ def __init__(self,
self.sliding_window = self.runner.sliding_window
self.block_size = self.runner.block_size
self.enable_lora = self.runner.lora_config is not None
self.multi_modal_input_mapper = self.runner.multi_modal_input_mapper
self.finished_requests_ids = finished_requests_ids
self.decode_only = True
self.is_encoder_decoder = self.runner.model_config.is_encoder_decoder
Expand Down Expand Up @@ -786,23 +785,15 @@ def _compute_lora_input(self, inter_data: InterDataForSeqGroup,
def _compute_multi_modal_input(self, inter_data: InterDataForSeqGroup,
seq_group_metadata: SequenceGroupMetadata):
"""If multi-modal data is given, add it to the input."""
# NOTE: mm_data only includes the subset of multi-modal items that
# NOTE: mm_kwargs only includes the subset of multi-modal items that
# intersect with the current prefill positions.
positions = inter_data.input_positions[0]
mm_data, placeholder_maps = MultiModalPlaceholderMap.from_seq_group(
mm_kwargs, placeholder_maps = MultiModalPlaceholderMap.from_seq_group(
seq_group_metadata,
range(positions[0], positions[0] + len(positions)))
if not mm_data:
if not mm_kwargs:
return

if self.runner.mm_registry.has_processor(self.runner.model_config):
mm_kwargs = mm_data
else:
mm_kwargs = self.multi_modal_input_mapper(
mm_data,
seq_group_metadata.mm_processor_kwargs,
)

inter_data.multi_modal_kwargs = mm_kwargs
inter_data.multi_modal_placeholder_maps = placeholder_maps

Expand Down Expand Up @@ -918,9 +909,6 @@ def __init__(
# Multi-modal data support
self.input_registry = input_registry
self.mm_registry = mm_registry
self.multi_modal_input_mapper = mm_registry \
.create_input_mapper(model_config)
self.mm_registry.init_mm_limits_per_prompt(self.model_config)

# Lazy initialization
self.model: nn.Module # Set after load_model
Expand Down Expand Up @@ -1116,8 +1104,8 @@ def profile_run(self) -> None:

dummy_data = self.input_registry \
.dummy_data_for_profiling(self.model_config,
seq_len,
self.mm_registry)
seq_len,
self.mm_registry)

seq = SequenceGroupMetadata(
request_id=str(group_id),
Expand Down