[Core] Cast multimodal input in hf processor (#18862)

Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
This commit is contained in:
Lukas Geiger
2025-06-03 20:24:56 -07:00
committed by GitHub
parent 4555143ea7
commit 1409ef9134
15 changed files with 25 additions and 25 deletions
-1
View File
@@ -300,7 +300,6 @@ class CPUEncoderDecoderModelRunner(
model_input.encoder_input_positions,
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
"intermediate_tensors":
-1
View File
@@ -630,7 +630,6 @@ class CPUModelRunner(CPUModelRunnerBase[ModelInputForCPUWithSamplingMetadata]):
if model_input.multi_modal_kwargs is not None:
multimodal_kwargs = MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
)
execute_model_kwargs = {}
-1
View File
@@ -53,7 +53,6 @@ class CPUPoolingModelRunner(
model_input.input_positions,
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
**cross_enc_kwargs,
-1
View File
@@ -205,7 +205,6 @@ class EncoderDecoderModelRunner(GPUModelRunnerBase[EncoderDecoderModelInput]):
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(
multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
),
**seqlen_agnostic_kwargs,
-1
View File
@@ -1848,7 +1848,6 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(
multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
),
**seqlen_agnostic_kwargs,
@@ -73,7 +73,6 @@ class MultiStepNeuronModelRunner(NeuronModelRunner):
input_block_ids=model_input.input_block_ids,
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)
@@ -52,7 +52,6 @@ class MultiStepNeuronxDistributedModelRunner(NeuronxDistributedModelRunner):
sampling_params=sampling_params,
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)
-2
View File
@@ -395,7 +395,6 @@ class NeuronModelRunner(ModelRunnerBase[ModelInputForNeuron]):
adapter_ids=model_input.adapter_ids,
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)
@@ -408,7 +407,6 @@ class NeuronModelRunner(ModelRunnerBase[ModelInputForNeuron]):
input_block_ids=model_input.input_block_ids,
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)
-1
View File
@@ -122,7 +122,6 @@ class PoolingModelRunner(
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(
multi_modal_kwargs,
dtype=self.model_config.dtype,
device=self.device,
),
**cross_enc_kwargs,
-1
View File
@@ -565,7 +565,6 @@ class XPUModelRunner(ModelRunnerBase[ModelInputForXPUWithSamplingMetadata]):
intermediate_tensors=intermediate_tensors,
**MultiModalKwargs.as_kwargs(
model_input.multi_modal_kwargs or {},
dtype=self.model_config.dtype,
device=self.device,
),
)