[Misc] Fix warnings for mistral model (#23552)

Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Signed-off-by: Jiangyun Zhu <riverclouds.zhu@qq.com>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
This commit is contained in:
Jiangyun Zhu
2025-08-29 06:58:48 +00:00
committed by GitHub
co-authored by Patrick von Platen
parent 2d0afcc9dc
commit 885ca6d31d
3 changed files with 31 additions and 23 deletions
+7 -5
View File
@@ -15,7 +15,7 @@ from mistral_common.protocol.instruct.messages import (ImageChunk, TextChunk,
from mistral_common.protocol.instruct.request import ChatCompletionRequest
from mistral_common.tokens.tokenizers.multimodal import ImageEncoder
from PIL import Image
from transformers import PixtralVisionConfig, TensorType
from transformers import BatchFeature, PixtralVisionConfig, TensorType
from transformers.image_utils import ImageInput
from transformers.models.pixtral.image_processing_pixtral import (
_num_image_tokens as _get_pixtral_hf_num_image_tokens)
@@ -163,10 +163,12 @@ class PixtralProcessorAdapter:
images_processed.append(image_processed)
images_tokens.append(image_tokens)
return {
"input_ids": torch.cat(images_tokens)[None].expand(len(text), -1),
"images": images_processed,
}
return BatchFeature({
"input_ids":
torch.cat(images_tokens)[None].expand(len(text), -1),
"images":
images_processed,
})
class PixtralProcessingInfo(BaseProcessingInfo):
+7 -5
View File
@@ -17,7 +17,7 @@ from mistral_common.protocol.instruct.messages import (AudioChunk, RawAudio,
from mistral_common.protocol.instruct.request import ChatCompletionRequest
from mistral_common.protocol.transcription.request import TranscriptionRequest
from mistral_common.tokens.tokenizers.audio import Audio, AudioEncoder
from transformers import TensorType, WhisperConfig
from transformers import BatchFeature, TensorType, WhisperConfig
from transformers.tokenization_utils_base import TextInput
from vllm.config import ModelConfig, SpeechToTextConfig, VllmConfig
@@ -156,10 +156,12 @@ class VoxtralProcessorAdapter:
audios_tokens.append(torch.tensor(audio_tokens))
audios_processed.append(torch.tensor(audio))
return {
"input_ids": torch.cat(audios_tokens)[None].expand(len(text), -1),
"audio_arrays": audios_processed,
}
return BatchFeature({
"input_ids":
torch.cat(audios_tokens)[None].expand(len(text), -1),
"audio_arrays":
audios_processed,
})
class VoxtralProcessingInfo(BaseProcessingInfo):