Remove unused kwargs from model definitions (#13555)
This commit is contained in:
@@ -4,8 +4,8 @@
|
||||
"""PyTorch Ultravox model."""
|
||||
import math
|
||||
from functools import cached_property
|
||||
from typing import (Any, Iterable, List, Literal, Mapping, Optional, Set,
|
||||
Tuple, TypedDict, Union)
|
||||
from typing import (Any, Iterable, Literal, Mapping, Optional, Set, Tuple,
|
||||
TypedDict, Union)
|
||||
|
||||
import torch
|
||||
import torch.utils.checkpoint
|
||||
@@ -16,8 +16,8 @@ from transformers.models.whisper import WhisperFeatureExtractor
|
||||
from transformers.models.whisper.modeling_whisper import WhisperEncoder
|
||||
|
||||
from vllm import envs
|
||||
from vllm.attention import AttentionMetadata
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.forward_context import get_forward_context
|
||||
from vllm.model_executor.layers.activation import MulAndSilu, get_act_fn
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
|
||||
@@ -495,13 +495,13 @@ class UltravoxModel(nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA):
|
||||
self,
|
||||
input_ids: torch.Tensor,
|
||||
multimodal_embeddings: Optional[NestedTensors] = None,
|
||||
attn_metadata: Optional[AttentionMetadata] = None,
|
||||
) -> torch.Tensor:
|
||||
inputs_embeds = self.language_model.get_input_embeddings(input_ids)
|
||||
if multimodal_embeddings is not None:
|
||||
|
||||
# TODO(ywang96): remove this block after v0 is deprecated.
|
||||
if not envs.VLLM_USE_V1:
|
||||
attn_metadata = get_forward_context().attn_metadata
|
||||
merge_multimodal_embeddings_from_map(
|
||||
inputs_embeds, multimodal_embeddings,
|
||||
attn_metadata.multi_modal_placeholder_index_maps["audio"])
|
||||
@@ -514,8 +514,6 @@ class UltravoxModel(nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA):
|
||||
def forward(self,
|
||||
input_ids: torch.Tensor,
|
||||
positions: torch.Tensor,
|
||||
kv_caches: List[torch.Tensor],
|
||||
attn_metadata: AttentionMetadata,
|
||||
intermediate_tensors: Optional[torch.Tensor] = None,
|
||||
inputs_embeds: Optional[torch.Tensor] = None,
|
||||
**kwargs) -> Union[torch.Tensor, IntermediateTensors]:
|
||||
@@ -540,17 +538,12 @@ class UltravoxModel(nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA):
|
||||
elif inputs_embeds is None:
|
||||
multimodal_embeddings = self.get_multimodal_embeddings(**kwargs)
|
||||
|
||||
# TODO(ywang96): remove attn_metadata from get_input_embeddings
|
||||
# after v0 is deprecated
|
||||
inputs_embeds = self.get_input_embeddings(input_ids,
|
||||
multimodal_embeddings,
|
||||
attn_metadata)
|
||||
multimodal_embeddings)
|
||||
input_ids = None
|
||||
|
||||
hidden_states = self.language_model.model(input_ids,
|
||||
positions,
|
||||
kv_caches,
|
||||
attn_metadata,
|
||||
intermediate_tensors,
|
||||
inputs_embeds=inputs_embeds)
|
||||
return hidden_states
|
||||
|
||||
Reference in New Issue
Block a user