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
File size: 4,298 Bytes
e3f7919 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 | {
"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"
}
|