Aller au contenu
ModelPrésentéExterne
Télécharger

Qwen35 9b Omni Kaggle Build

Qwen3.5-9B native multimodal + Whisper audio branch + Jinja thinking template.

Sourcehuggingface.co/Hellohihihih/qwen35-9b-omni-kaggle-build/tree/011a46256568cdc29e09e49377d7dd6206985f43
TypeOtherModèle de baseQwen/Qwen3.5-9BVersionRevision 011a462CClasse

Source du modèle

Type
Other
Taille du fichier
23,1 MB
Modèle de base
Qwen/Qwen3.5-9B
Version
vRevision 011a462
Mis à jour
17 oct. 2026
Téléchargements
0
Source
Hugging Face

Description de la source

Qwen35 Omni Kaggle Build

Qwen3.5-9B native multimodal + Whisper audio branch + Jinja thinking template.

Runtime safety:

  • greedy decode for audio branch
  • audio/text embeddings use nan_to_num + clamp before generation

Stock LM Studio requires llama.cpp qwen35_omni backend support for the custom audio branch.

Sources

1 source
Hugging FacePrincipalVersion 011a46256568cdc29e09e49377d7dd6206985f43
huggingface.co/Hellohihihih/qwen35-9b-omni-kaggle-build/tree/011a46256568cdc29e09e49377d7dd6206985f43

Vérifié 17 août

Artefacts du modèle

1 artefact

audio_projector.safetensors

safetensors · 23,1 MB · SHA-256 899b7b40414b…16b2 · Hugging Face

Télécharger

Extraits de sources

2 extraits
Readme17 aoûthuggingface.co/Hellohihihih/qwen35-9b-omni-kaggle-build

--- library_name: transformers base_model: - Qwen/Qwen3.5-9B tags: - qwen3.5 - omni - audio - video - image - korean - english - thinking pipeline_tag: any-to-any --- # Qwen35 Omni Kaggle Build Qwen3.5-9B native multimodal + Whisper audio branch + Jinja thinking template. Runtime safety: - greedy decode for audio branch - audio/text embeddings use nan_to_num + clamp before generation Stock LM Studio requires llama.cpp qwen35_omni backend support for the custom audio branch.

Note17 aoûthuggingface.co/Hellohihihih/qwen35-9b-omni-kaggle-build

Source context: 0 downloads · 1 likes · Pipeline any-to-any · Library transformers · Repo Hellohihihih/qwen35-9b-omni-kaggle-build