MangaliCa is the first publicly available Hungarian–English bilingual vision–language model designed for image captioning and image–text retrieval. The model is built on the CoCa (Contrastive Captioner) framework and...
Model source
Source excerpt
MangaliCa is the first publicly available Hungarian–English bilingual vision–language model designed for image captioning and image–text retrieval. The model is built on the CoCa (Contrastive Captioner) framework and...
Sources
1 sourceVerified Sep 17
Model artifacts
64 artifactsSource excerpts
2 excerptscheckpoints/checkpoint-101376-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 6eab94bb6550…433a · Hugging Face
Downloadcheckpoints/checkpoint-101376-gbc/scaler.pt
pt · 988 B · SHA-256 5f1631216cf7…ec43 · Hugging Face
Downloadcheckpoints/checkpoint-1024/model.safetensors
safetensors · 2.96 GB · SHA-256 606d83ade37f…a041 · Hugging Face
Downloadcheckpoints/checkpoint-10240/model.safetensors
safetensors · 2.96 GB · SHA-256 8d21b1a7297e…fa37 · Hugging Face
Downloadcheckpoints/checkpoint-102400-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 0db6dc2d7b19…1cda · Hugging Face
Downloadcheckpoints/checkpoint-102400-gbc/scaler.pt
pt · 988 B · SHA-256 ec362cc53202…3684 · Hugging Face
Downloadcheckpoints/checkpoint-103424-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 4c29c9f9a4cb…c7ea · Hugging Face
Downloadcheckpoints/checkpoint-103424-gbc/scaler.pt
pt · 988 B · SHA-256 70e362b650e8…a994 · Hugging Face
Downloadcheckpoints/checkpoint-104704-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 09e76b1435b3…5720 · Hugging Face
Downloadcheckpoints/checkpoint-104704-gbc/scaler.pt
pt · 988 B · SHA-256 00e1411c2382…e922 · Hugging Face
Downloadcheckpoints/checkpoint-105472-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 540ce641cfb7…642b · Hugging Face
Downloadcheckpoints/checkpoint-105472-gbc/scaler.pt
pt · 988 B · SHA-256 f183c8ebab3f…36aa · Hugging Face
Downloadcheckpoints/checkpoint-106496-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 8bc4f7137687…e14f · Hugging Face
Downloadcheckpoints/checkpoint-106496-gbc/scaler.pt
pt · 988 B · SHA-256 ec8baa9db52b…cd3f · Hugging Face
Downloadcheckpoints/checkpoint-107520-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 2703c1925000…3bfa · Hugging Face
Downloadcheckpoints/checkpoint-107520-gbc/scaler.pt
pt · 988 B · SHA-256 015db6c8ee20…fa06 · Hugging Face
Downloadcheckpoints/checkpoint-108544-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 79390a4d581f…7f64 · Hugging Face
Downloadcheckpoints/checkpoint-108544-gbc/scaler.pt
pt · 988 B · SHA-256 bce2a5da7dae…9295 · Hugging Face
Downloadcheckpoints/checkpoint-109568-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 f48ea0899d0c…bffb · Hugging Face
Downloadcheckpoints/checkpoint-109568-gbc/scaler.pt
pt · 988 B · SHA-256 b14de7a0ca92…1ed4 · Hugging Face
Downloadcheckpoints/checkpoint-110592-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 79fc33f0e8a5…e92a · Hugging Face
Downloadcheckpoints/checkpoint-110592-gbc/scaler.pt
pt · 988 B · SHA-256 65576acb0800…00dc · Hugging Face
Downloadcheckpoints/checkpoint-111616-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 046a46befae2…744c · Hugging Face
Downloadcheckpoints/checkpoint-111616-gbc/scaler.pt
pt · 988 B · SHA-256 4cb44dfbd014…f07b · Hugging Face
Downloadcheckpoints/checkpoint-11264/model.safetensors
safetensors · 2.96 GB · SHA-256 c5ddd6894ebd…e88d · Hugging Face
Downloadcheckpoints/checkpoint-112640-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 db3f48f31984…2e28 · Hugging Face
Downloadcheckpoints/checkpoint-112640-gbc/scaler.pt
pt · 988 B · SHA-256 c94e5dcb862e…fe2a · Hugging Face
Downloadcheckpoints/checkpoint-113664-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 eab8675aa035…f6c0 · Hugging Face
Downloadcheckpoints/checkpoint-113664-gbc/scaler.pt
pt · 988 B · SHA-256 31a3f2013675…264d · Hugging Face
Downloadcheckpoints/checkpoint-114688-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 d77217726581…8a25 · Hugging Face
Downloadcheckpoints/checkpoint-114688-gbc/scaler.pt
pt · 988 B · SHA-256 3a4f663e129a…854c · Hugging Face
Downloadcheckpoints/checkpoint-115712-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 d3c314218a74…175c · Hugging Face
Downloadcheckpoints/checkpoint-115712-gbc/scaler.pt
pt · 988 B · SHA-256 1634d80c1cff…8de2 · Hugging Face
Downloadcheckpoints/checkpoint-116736-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 0798787e9e09…4a3b · Hugging Face
Downloadcheckpoints/checkpoint-116736-gbc/scaler.pt
pt · 988 B · SHA-256 79c11ae5dd5d…b02c · Hugging Face
Downloadcheckpoints/checkpoint-117760-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 8699b870b439…87b9 · Hugging Face
Downloadcheckpoints/checkpoint-117760-gbc/scaler.pt
pt · 988 B · SHA-256 6a90c43f637e…ff5b · Hugging Face
Downloadcheckpoints/checkpoint-118784-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 08e4c73526f5…39ad · Hugging Face
Downloadcheckpoints/checkpoint-118784-gbc/scaler.pt
pt · 988 B · SHA-256 ddb6ccce5d82…a0cd · Hugging Face
Downloadcheckpoints/checkpoint-119808-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 b2cf8af767c3…ad79 · Hugging Face
Downloadcheckpoints/checkpoint-119808-gbc/scaler.pt
pt · 988 B · SHA-256 b5922deaf0da…6073 · Hugging Face
Downloadcheckpoints/checkpoint-120832-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 2094b36f0947…6715 · Hugging Face
Downloadcheckpoints/checkpoint-120832-gbc/scaler.pt
pt · 988 B · SHA-256 5a46d184f876…fc74 · Hugging Face
Downloadcheckpoints/checkpoint-121856-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 e469f0d7b0df…6dfe · Hugging Face
Downloadcheckpoints/checkpoint-121856-gbc/scaler.pt
pt · 988 B · SHA-256 a307ea32741c…0400 · Hugging Face
Downloadcheckpoints/checkpoint-12288/model.safetensors
safetensors · 2.96 GB · SHA-256 5981e77762ff…184b · Hugging Face
Downloadcheckpoints/checkpoint-122880-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 6574130ee5c6…651e · Hugging Face
Downloadcheckpoints/checkpoint-122880-gbc/scaler.pt
pt · 988 B · SHA-256 bbb3f2c7a115…0c6f · Hugging Face
Downloadcheckpoints/checkpoint-123904-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 e49087e9b229…7a8c · Hugging Face
Downloadcheckpoints/checkpoint-123904-gbc/scaler.pt
pt · 988 B · SHA-256 a9b21b00077c…cb94 · Hugging Face
Downloadcheckpoints/checkpoint-125184-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 bf08e38cddc2…b8c1 · Hugging Face
Downloadcheckpoints/checkpoint-125184-gbc/scaler.pt
pt · 988 B · SHA-256 b378331dcc4b…5975 · Hugging Face
Downloadcheckpoints/checkpoint-125952-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 6aef7d08312c…ecf7 · Hugging Face
Downloadcheckpoints/checkpoint-125952-gbc/scaler.pt
pt · 988 B · SHA-256 d3dc67c2d740…78b4 · Hugging Face
Downloadcheckpoints/checkpoint-126976-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 92b5adb5dff5…a288 · Hugging Face
Downloadcheckpoints/checkpoint-126976-gbc/scaler.pt
pt · 988 B · SHA-256 82b8d7f73ed9…aa2c · Hugging Face
Downloadcheckpoints/checkpoint-128000-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 8808c0ca9255…d009 · Hugging Face
Downloadcheckpoints/checkpoint-128000-gbc/scaler.pt
pt · 988 B · SHA-256 37869c8fad2a…ffb9 · Hugging Face
Downloadcheckpoints/checkpoint-129024-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 3f0602b92b98…0edc · Hugging Face
Downloadcheckpoints/checkpoint-129024-gbc/scaler.pt
pt · 988 B · SHA-256 17bc2ec4967f…0126 · Hugging Face
Downloadcheckpoints/checkpoint-130048-gbc/model.safetensors
safetensors · 2.96 GB · SHA-256 45743600dab5…7b90 · Hugging Face
Downloadcheckpoints/checkpoint-130048-gbc/scaler.pt
pt · 988 B · SHA-256 36801e333283…08d0 · Hugging Face
Download--- language: - en - hu license: cc-by-nc-sa-4.0 tags: - vision-language - image-captioning - image-text-retrieval - multilingual datasets: - Obscure-Entropy/MangaliCa_EN-HU metrics: - recall@k - mrr - ndcg pipeline_tag: image-to-text library_name: transformers model-index: - name: MangaliCa results: - task: type: image-text-retrieval dataset: type: image-text name: GBC-10M metrics: - name: R@1 type: R@1 value: 0.356 verified: false - task: type: image-text-retrieval dataset: type: image-text name: MS-COCO metrics: - name: R@1 type: R@1 value: 0.060 verified: false - task: type: image-text-retrieval dataset: type: image-text name: text-to-image-2M metrics: - name: R@1 type: R@1 value: 0.415 verified: false - task: type: image-text-retrieval dataset: type: image-text name: XM3600 metrics: - name: R@1 type: R@1 value: 0.113 verified: false --- # 🐗 MangaliCa: A Bilingual Vision–Language Model for Hungarian–English Image Captioning and Retrieval <div align="center"> <img src="./assets/mangalica.jpg" width="50%" /> </div> ## Model Description **MangaliCa** is the first publicly available Hungarian–English bilingual vision–language model designed for image captioning and image–text retrieval. The model is built on the CoCa (Contrastive Captioner) framework and jointly optimizes contrastive alignment and autoregressive caption generation across two languages. MangaliCa integrates: - CLIP ViT-L/14 as a frozen vision encoder - TinyLlama 1.1B as a text decoder, extended with cross-attention - Language-conditioning tokens for explicit bilingual control - LoRA adapters for memory-efficient fine-tuning > [!CAUTION] > The model was trained on a newly constructed **70M-sample Hungarian–English bilingual image–caption dataset**, the largest multimodal dataset involving Hungarian to date. --- ## Intended Uses ### Supported Tasks - **Image captioning** (Hungarian and English) - **Image–text retrieval** (text-to-image, image-to-text) - **Cross-lingual text–text retrieval** - Multimodal embedding extraction for RAG or search systems ### Example Use Cases - Hungarian-language multimodal assistants - Bilingual image search engines - Vision-enabled Retrieval-Augmented Generation (RAG) - Research on medium-resource multilingual vision–language learning --- ## 🪛 Model Architecture - **Vision Encoder**: CLIP ViT-L/14 (frozen,...
Source context: 0 downloads · 0 likes · Pipeline image-to-text · Library transformers · Repo Obscure-Entropy/MangaliCa