A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Docs] add __init__.py to vllm/model_executor/layers/quantization/compressed_tensors/transform (#24974)
Signed-off-by: samzong <samzong.lu@gmail.com>
samzong committed
Sep 20, 2025 at 02:32 UTC
138f0d1e752d3a35cd959f3df8bf00370a2ace7b
5 files changed
+5
-4
vllm/model_executor/layers/quantization/compressed_tensors/transform/__init__.py
vllm/model_executor/layers/quantization/compressed_tensors/transform/schemes/__init__.py
vllm/model_executor/models/blip2.py
+1
-1
@@ -680,7 +680,7 @@ class Blip2ForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP,
680
batch.
681
682
Info:
683
- [Blip2ImageInputs][]
683
+ [`Blip2ImageInputs`][vllm.model_executor.models.blip2.Blip2ImageInputs]
684
"""
685
686
if intermediate_tensors is not None:
vllm/model_executor/models/llava.py
+1
-1
@@ -737,7 +737,7 @@ class LlavaForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP):
737
inputs_embeds: Optional tensor of input embeddings.
738
739
Info:
740
- [LlavaImageInputs][]
740
+ [`LlavaImageInputs`][vllm.model_executor.models.llava.LlavaImageInputs]
741
"""
742
if intermediate_tensors is not None:
743
inputs_embeds = None
vllm/model_executor/models/llava_next.py
+3
-2
@@ -527,7 +527,8 @@ class LlavaNextForConditionalGeneration(nn.Module, SupportsMultiModal,
527
Unlike in LLaVA-1.5, the number of image tokens inputted to the language
528
model depends on the original size of the input image. Including the
529
original image token in the input, the required number of image tokens
530
- is given by [get_llava_next_image_feature_size][].
530
+ is given by [`LlavaNextProcessingInfo.get_num_image_tokens`][vllm.\
531
+model_executor.models.llava_next.LlavaNextProcessingInfo.get_num_image_tokens].
532
533
This way, the `positions` and `attn_metadata` are consistent
534
with the `input_ids`.
@@ -540,7 +541,7 @@ class LlavaNextForConditionalGeneration(nn.Module, SupportsMultiModal,
541
inputs_embeds: Optional tensor of input embeddings.
542
543
Info:
543
- [LlavaNextImageInputs][]
544
+ [`LlavaNextImageInputs`][vllm.model_executor.models.llava_next.LlavaNextImageInputs]
545
"""
546
if intermediate_tensors is not None:
547
inputs_embeds = None