Skip to content

[V1] Support Mistral3 in V1 #15950

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
Apr 2, 2025
Merged

[V1] Support Mistral3 in V1 #15950

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2157bef
[V1] Support Mistral3 in V1
mgoin Apr 2, 2025
7313922
Update docs
mgoin Apr 2, 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
2 changes: 1 addition & 1 deletion docs/source/models/supported_models.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -888,7 +888,7 @@ See [this page](#generative-models) for more information on how to use generativ
* `mistralai/Mistral-Small-3.1-24B-Instruct-2503`, etc.
*
* ✅︎
*
* ✅︎
- * `MllamaForConditionalGeneration`
* Llama 3.2
* T + I<sup>+</sup>
Expand Down
15 changes: 9 additions & 6 deletions vllm/model_executor/models/mistral3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,12 +31,12 @@
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
from vllm.sequence import IntermediateTensors

from .interfaces import (MultiModalEmbeddings, SupportsMultiModal, SupportsPP,
SupportsV0Only)
from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsPP
from .pixtral import PixtralHFEncoderInfo, PixtralHFVisionModel
from .utils import (AutoWeightsLoader, flatten_bn, init_vllm_registered_model,
maybe_prefix, merge_multimodal_embeddings)
from .vision import get_vision_encoder_info, select_patch_features
from .vision import (get_vision_encoder_info, scatter_patch_features,
select_patch_features)


class Mistral3ImagePixelInputs(TypedDict):
Expand Down Expand Up @@ -425,7 +425,7 @@ def init_vision_tower_for_llava(
info=_build_mistral3_info,
dummy_inputs=Mistral3DummyInputsBuilder)
class Mistral3ForConditionalGeneration(nn.Module, SupportsMultiModal,
SupportsPP, SupportsV0Only):
SupportsPP):

packed_modules_mapping = {
"qkv_proj": ["q_proj", "k_proj", "v_proj"],
Expand Down Expand Up @@ -518,7 +518,7 @@ def _parse_and_validate_image_input(
return Mistral3ImagePixelInputs(
type="pixel_values_pixtral",
pixel_values=flatten_bn(pixel_values),
embed_is_patch=embed_is_patch,
embed_is_patch=flatten_bn(embed_is_patch),
)

def _process_image_input(
Expand Down Expand Up @@ -557,7 +557,10 @@ def get_multimodal_embeddings(

vision_embeddings = self._process_image_input(image_input)

return vision_embeddings
return scatter_patch_features(
vision_embeddings,
image_input["embed_is_patch"],
)

def get_input_embeddings(
self,
Expand Down