Image-Text-to-Text
Transformers
Safetensors
Arabic
English
arabic
vision
image-captioning
image-understanding
siglip2
jais
Instructions to use dhadlab/dhad-vision with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use dhadlab/dhad-vision with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="dhadlab/dhad-vision")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("dhadlab/dhad-vision", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use dhadlab/dhad-vision with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "dhadlab/dhad-vision" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "dhadlab/dhad-vision", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/dhadlab/dhad-vision
- SGLang
How to use dhadlab/dhad-vision with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "dhadlab/dhad-vision" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "dhadlab/dhad-vision", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "dhadlab/dhad-vision" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "dhadlab/dhad-vision", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use dhadlab/dhad-vision with Docker Model Runner:
docker model run hf.co/dhadlab/dhad-vision
Download MANIFEST.json from dhadlab/dhad-vision: direct link, hf CLI and curl.
- Browser
- Download file 4.3 kB
-
https://huggingface.co/dhadlab/dhad-vision/resolve/main/MANIFEST.json
- Command line
-
hf download hf://dhadlab/dhad-vision/MANIFEST.json
-
curl -L -o MANIFEST.json https://huggingface.co/dhadlab/dhad-vision/resolve/main/MANIFEST.json
4.3 kB
| { | |
| "part": "vision-preview", | |
| "public_name": "dhad-1-vision-preview", | |
| "arabic": "رؤية", | |
| "release": "dhad-1", | |
| "status": "preview", | |
| "licence": "Apache-2.0", | |
| "internal_checkpoint": "brain2/vision/stage1/final", | |
| "copied": "2026-11-15", | |
| "base_model": { | |
| "id": "inception42/Jais-3-0.9B (Hugging Face; README link; local copy brain2/jais3_09b)", | |
| "licence": "Apache-2.0", | |
| "trunk_parameters": 1078285824 | |
| }, | |
| "mix": "mixes/dhad-1/vision-preview.json", | |
| "parameters": { | |
| "connector": 7080961, | |
| "markers": 3072, | |
| "connector_plus_markers": 7084033, | |
| "connector_tensors": 6, | |
| "encoder_siglip2_vision_tower_used": 85843200, | |
| "encoder_vision_tower_with_pooling_head": 92930304, | |
| "encoder_pooling_head_unused": 7087104, | |
| "siglip2_text_tower_unused": 282303744, | |
| "siglip2_total_file": 375234050, | |
| "architecture": "SigLIP2-Base NaFlex vision tower (frozen, 384x384, 576 patches x 768, last_hidden_state after post-layernorm, pooling head not used) -> pixel-shuffle 2x2 -> Linear 3072->1536 -> GELU -> Linear 1536->1536 -> L2 x learnable scale -> 144 soft tokens (64 in low mode) + 2 marker vectors, into the frozen Jais trunk" | |
| }, | |
| "encoder": { | |
| "id": "google/siglip2-base-patch16-naflex family (local copy brain2/vision/siglip2; config says Siglip2Model; exact repo id not recorded in the folder)", | |
| "licence": "Apache-2.0 (card front matter, brain2/vision/siglip2/README.md)", | |
| "file": "brain2/vision/siglip2/model.safetensors", | |
| "sha256": "ac5f28bbdf92c0c1696ccbd3ce716426049cd67ad8045b66d0d938b0f9c8bbec", | |
| "bytes": 1500985224, | |
| "not_reuploaded": true | |
| }, | |
| "stage_included": "stage 1 (connector + markers; frozen trunk). No vision LoRA in this folder: the stage-2 vision LoRA is unfinished and lives in stage2-step675/ (extra).", | |
| "stage2_step675": { | |
| "status": "UNFINISHED (step 675 of 1,435, gates never run)", | |
| "folder": "stage2-step675", | |
| "files": [ | |
| { | |
| "path": "connector.pt", | |
| "source": "brain2/vision/stage2/last/connector.pt", | |
| "sha256": "ed6bdc9f72080a6dd69546c0cc722b6157f9bfbbb640909963ab90b2f7ea009d", | |
| "bytes": 28339181, | |
| "source_sha256_equal": true | |
| }, | |
| { | |
| "path": "vision_lora/adapter_config.json", | |
| "source": "brain2/vision/stage2/last/vision_lora/adapter_config.json", | |
| "sha256": "8d4bbde6cfddaee01fe287152da3eb2a443abed3b45d59bd834ad84d0bcc6c7e", | |
| "bytes": 1200, | |
| "source_sha256_equal": true | |
| }, | |
| { | |
| "path": "vision_lora/adapter_model.safetensors", | |
| "source": "brain2/vision/stage2/last/vision_lora/adapter_model.safetensors", | |
| "sha256": "b2d84e923c7fb3e9a7858eb629afc215958af546ca9a2c811c3a7f89b5d243e8", | |
| "bytes": 111987648, | |
| "source_sha256_equal": true | |
| }, | |
| { | |
| "path": "UNFINISHED.md", | |
| "sha256": "ff015e936d4bea5b5deee440d7aea7185217aff05db8986377ea8cfd3446db88", | |
| "bytes": 537 | |
| } | |
| ], | |
| "vision_lora": { | |
| "tensors": 392, | |
| "parameters": 27983872, | |
| "rank": 32, | |
| "alpha": 64, | |
| "dropout": 0.05, | |
| "target_modules": [ | |
| "q_proj", | |
| "v_proj", | |
| "k_proj", | |
| "up_proj", | |
| "gate_proj", | |
| "down_proj", | |
| "o_proj" | |
| ], | |
| "dtype": "fp32 on disk (served in bf16)", | |
| "base_model_name_or_path_recorded": "/home/shami/projects/tashih/brain2/jais3_09b", | |
| "peft_version": "0.21.2" | |
| }, | |
| "connector_parameters": 7080961, | |
| "marker_parameters": 3072, | |
| "marker_file": "stage2-step675/UNFINISHED.md" | |
| }, | |
| "files": [ | |
| { | |
| "path": "connector.pt", | |
| "source": "brain2/vision/stage1/final/connector.pt", | |
| "sha256": "82c611cadf8e694d23b5edea89297a5647c61417f45e11a168f5d9cafc63dd22", | |
| "bytes": 28339217, | |
| "source_sha256_equal": true | |
| } | |
| ], | |
| "bytes_total": 168667783, | |
| "gate_results": [ | |
| { | |
| "file": "logs/vision_stage1.log", | |
| "note": "log with the PROBE lines" | |
| } | |
| ], | |
| "trained_date": "2026-11-14 18:31 (checkpoint file mtime)", | |
| "training_data": [ | |
| { | |
| "dataset": "dhad-vision-captions-v1", | |
| "train": 29800, | |
| "test": 200, | |
| "features_excluded_gb": 25 | |
| } | |
| ], | |
| "training_data_note": "see the mix file for the init chain, hyper-parameters, source shares and licence flags" | |
| } | |