Skip to content

[Core] Automatically cast multi-modal input dtype #18756

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 5 commits into from
May 27, 2025
Merged

[Core] Automatically cast multi-modal input dtype #18756

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
01b9fe6
[Core] Automatically cast multi-modal input dtype
DarkLight1337 May 27, 2025
da2ff28
Add comment
DarkLight1337 May 27, 2025
b46e994
Fix type
DarkLight1337 May 27, 2025
8da6864
Remove dtype conversion
DarkLight1337 May 27, 2025
b686bd9
Update
DarkLight1337 May 27, 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
4 changes: 1 addition & 3 deletions vllm/model_executor/models/deepseek_vl2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -210,9 +210,7 @@ def _call_hf_processor(
dict(prompt=prompt, **mm_data),
mm_kwargs,
)
target_dtype = self.info.ctx.model_config.dtype
pixel_values = processed_outputs.pop("pixel_values").to(
target_dtype)
pixel_values = processed_outputs["pixel_values"]
# split pixel values into patches corresponding to each image
images_spatial_crop = processed_outputs["images_spatial_crop"]
patches_per_image = [
Expand Down
8 changes: 7 additions & 1 deletion vllm/multimodal/inputs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -731,11 +731,17 @@ def as_kwargs(
batched_inputs: BatchedTensorInputs,
*,
device: torch.types.Device,
dtype: Optional[torch.dtype] = None,
) -> BatchedTensorInputs:
json_inputs = cast(JSONTree[torch.Tensor], batched_inputs)

json_mapped = json_map_leaves(
lambda x: x.to(device, non_blocking=True),
lambda x: x.to(
device=device,
# This mimics the behavior of transformers.BatchFeature
dtype=dtype if x.is_floating_point() else None,
non_blocking=True,
),
json_inputs,
)

Expand Down
7 changes: 5 additions & 2 deletions vllm/spec_decode/draft_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -294,8 +294,11 @@ def execute_model(
inputs_embeds=None,
positions=model_input.input_positions,
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(multi_modal_kwargs,
device=self.device),
**MultiModalKwargs.as_kwargs(
multi_modal_kwargs,
dtype=self.model_runner.model_config.dtype,
device=self.device,
),
**model_execute_kwargs,
)

Expand Down
12 changes: 9 additions & 3 deletions vllm/v1/worker/gpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -929,8 +929,11 @@ def _execute_mm_encoder(self, scheduler_output: "SchedulerOutput"):
encoder_outputs = []
for grouped_mm_inputs in grouped_mm_inputs_list:
batched_mm_inputs = MultiModalKwargs.batch(grouped_mm_inputs)
batched_mm_inputs = MultiModalKwargs.as_kwargs(batched_mm_inputs,
device=self.device)
batched_mm_inputs = MultiModalKwargs.as_kwargs(
batched_mm_inputs,
dtype=self.model_config.dtype,
device=self.device,
)

# Run the encoder.
# `curr_group_outputs` is either of the following:
Expand Down Expand Up @@ -1874,7 +1877,10 @@ def profile_run(self) -> None:
batched_dummy_mm_inputs = MultiModalKwargs.batch(
[dummy_mm_kwargs] * max_num_mm_items)
batched_dummy_mm_inputs = MultiModalKwargs.as_kwargs(
batched_dummy_mm_inputs, device=self.device)
batched_dummy_mm_inputs,
dtype=self.model_config.dtype,
device=self.device,
)

# Run multimodal encoder.
dummy_encoder_outputs = self.model.get_multimodal_embeddings(
Expand Down
14 changes: 10 additions & 4 deletions vllm/v1/worker/tpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -652,8 +652,11 @@ def _execute_mm_encoder(self, scheduler_output: "SchedulerOutput"):
encoder_outputs = []
for grouped_mm_inputs in grouped_mm_inputs_list:
batched_mm_inputs = MultiModalKwargs.batch(grouped_mm_inputs)
batched_mm_inputs = MultiModalKwargs.as_kwargs(batched_mm_inputs,
device=self.device)
batched_mm_inputs = MultiModalKwargs.as_kwargs(
batched_mm_inputs,
dtype=self.model_config.dtype,
device=self.device,
)

# Run the encoder.
# `curr_group_outputs` is either of the following:
Expand Down Expand Up @@ -1435,8 +1438,11 @@ def _get_mm_dummy_batch(self, modality: str,

batched_dummy_mm_inputs = MultiModalKwargs.batch([dummy_mm_kwargs] *
batch_size)
return MultiModalKwargs.as_kwargs(batched_dummy_mm_inputs,
device=self.device)
return MultiModalKwargs.as_kwargs(
batched_dummy_mm_inputs,
dtype=self.model_config.dtype,
device=self.device,
)


def _get_req_paddings(min_req_size: int, max_req_size: int) -> list[int]:
Expand Down
7 changes: 5 additions & 2 deletions vllm/worker/cpu_enc_dec_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -297,8 +297,11 @@ def execute_model(
model_input.encoder_input_tokens,
"encoder_positions":
model_input.encoder_input_positions,
**MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs or {},
device=self.device),
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
"intermediate_tensors":
intermediate_tensors,
}
Expand Down
5 changes: 4 additions & 1 deletion vllm/worker/cpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -628,7 +628,10 @@ def execute_model(
multimodal_kwargs = {}
if model_input.multi_modal_kwargs is not None:
multimodal_kwargs = MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs, device=self.device)
model_input.multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
)
execute_model_kwargs = {}
if previous_hidden_states is not None:
execute_model_kwargs.update(
Expand Down
7 changes: 5 additions & 2 deletions vllm/worker/cpu_pooling_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,8 +50,11 @@ def execute_model(
model_input.input_tokens,
"positions":
model_input.input_positions,
**MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs or {},
device=self.device),
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
**cross_enc_kwargs,
"intermediate_tensors":
intermediate_tensors,
Expand Down
10 changes: 7 additions & 3 deletions vllm/worker/enc_dec_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -202,9 +202,13 @@ def execute_model(
encoder_input_ids=model_input.encoder_input_tokens,
encoder_positions=model_input.encoder_input_positions,
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(multi_modal_kwargs,
device=self.device),
**seqlen_agnostic_kwargs)
**MultiModalKwargs.as_kwargs(
multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
),
**seqlen_agnostic_kwargs,
)

logits = self.model.compute_logits(hidden_or_intermediate_states,
model_input.sampling_metadata)
Expand Down
7 changes: 5 additions & 2 deletions vllm/worker/model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1845,8 +1845,11 @@ def execute_model(
inputs_embeds=model_input.inputs_embeds,
positions=model_input.input_positions,
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(multi_modal_kwargs,
device=self.device),
**MultiModalKwargs.as_kwargs(
multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
),
**seqlen_agnostic_kwargs,
**model_kwargs,
)
Expand Down
7 changes: 5 additions & 2 deletions vllm/worker/multi_step_neuron_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,8 +70,11 @@ def execute_model(
input_ids=model_input.input_tokens,
positions=model_input.input_positions,
input_block_ids=model_input.input_block_ids,
**MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs or {},
device=self.device),
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)

output = self.model.sample(
Expand Down
7 changes: 5 additions & 2 deletions vllm/worker/multi_step_neuronx_distributed_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,8 +49,11 @@ def execute_model(
positions=model_input.input_positions,
input_block_ids=model_input.input_block_ids,
sampling_params=sampling_params,
**MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs or {},
device=self.device),
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)

output = self.model.sample(
Expand Down
16 changes: 10 additions & 6 deletions vllm/worker/neuron_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -378,9 +378,11 @@ def execute_model(
positions=model_input.input_positions,
input_block_ids=model_input.input_block_ids,
sampling_params=sampling_params,
**MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs
or {},
device=self.device),
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)
elif current_platform.use_transformers_neuronx():
# [TODO] validate on-device sampling
Expand All @@ -389,9 +391,11 @@ def execute_model(
input_ids=model_input.input_tokens,
positions=model_input.input_positions,
input_block_ids=model_input.input_block_ids,
**MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs
or {},
device=self.device),
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)

# Compute the logits only if the on-device sampling is turned off as
Expand Down
10 changes: 7 additions & 3 deletions vllm/worker/pooling_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,10 +119,14 @@ def execute_model(
input_ids=model_input.input_tokens,
positions=model_input.input_positions,
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(multi_modal_kwargs,
device=self.device),
**MultiModalKwargs.as_kwargs(
multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
),
**cross_enc_kwargs,
**seqlen_agnostic_kwargs)
**seqlen_agnostic_kwargs,
)

if (self.observability_config is not None
and self.observability_config.collect_model_forward_time):
Expand Down
9 changes: 6 additions & 3 deletions vllm/worker/xpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -562,9 +562,12 @@ def execute_model(
input_ids=model_input.input_tokens,
positions=model_input.input_positions,
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(model_input.multi_modal_kwargs
or {},
device=self.device))
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)
# Compute the logits in the last pipeline stage.
if not get_pp_group().is_last_rank:
return hidden_or_intermediate_states
Expand Down