Image-Text-to-Text
PEFT
Safetensors
MLX
English
typed-decisions
jev
decision-model
vision-language
calibrated-probabilities
lora
Instructions to use mohit67890/imajev-9b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use mohit67890/imajev-9b with PEFT:
Task type is invalid.
- MLX
How to use mohit67890/imajev-9b with MLX:
# Make sure mlx-vlm is installed # pip install --upgrade mlx-vlm from mlx_vlm import load, generate from mlx_vlm.prompt_utils import apply_chat_template from mlx_vlm.utils import load_config # Load the model model, processor = load("mohit67890/imajev-9b") config = load_config("mohit67890/imajev-9b") # Prepare input image = ["http://images.cocodataset.org/val2017/000000039769.jpg"] prompt = "Describe this image." # Apply chat template formatted_prompt = apply_chat_template( processor, config, prompt, num_images=1 ) # Generate output output = generate(model, processor, formatted_prompt, image) print(output) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download calibration.json from mohit67890/imajev-9b: direct link, hf CLI and curl.
- Browser
- Download file 1.57 kB
-
https://huggingface.co/mohit67890/imajev-9b/resolve/main/calibration.json
- Command line
-
hf download hf://mohit67890/imajev-9b/calibration.json
-
curl -L -o calibration.json https://huggingface.co/mohit67890/imajev-9b/resolve/main/calibration.json
1.57 kB
| { | |
| "schema_version": "1.1", | |
| "calibration_version": "imajev-9b-1.0", | |
| "temperatures": { | |
| "boolean:2": 1.748, | |
| "boolean:3-5": 1.748, | |
| "boolean:6-10": 1.748, | |
| "boolean:11-25": 1.748, | |
| "boolean:26-254": 1.748, | |
| "choice:2": 1.748, | |
| "choice:3-5": 1.748, | |
| "choice:6-10": 1.748, | |
| "choice:11-25": 1.748, | |
| "choice:26-254": 1.748, | |
| "ordinal:2": 1.748, | |
| "ordinal:3-5": 1.748, | |
| "ordinal:6-10": 1.748, | |
| "ordinal:11-25": 1.748, | |
| "ordinal:26-254": 1.748 | |
| }, | |
| "unknown_offsets": {}, | |
| "fit": { | |
| "set": "data/manifests/decision-p2b-jevstyle-dev.jsonl (150 authored JevBench-style items, no JevBench items)", | |
| "method": "single temperature by NLL, applied to every (type, option-count) bucket (phase-2b policy)", | |
| "adapter": "phase-2c soup50 = 0.5 x phase-2b 9b adapter + 0.5 x phase-2c 9b best (LoRA + readout averaged)", | |
| "temperature": 1.748, | |
| "nll_raw": 0.7615848725618098, | |
| "nll_calibrated": 0.6568009539784012, | |
| "note": "served JevBench hard with this file: 69.4 / ECE 0.092 (rot4 + cal), 69.4 / 0.090 without rotations; the pod's phase-2c fold gives T~1 and does nothing on hard; offsets 0 (unknown handling gated separately)", | |
| "date": "2026-09-24", | |
| "internal_version": "p2c-9b-soup50-authored-dev" | |
| }, | |
| "counts": { | |
| "boolean:2": 150, | |
| "boolean:3-5": 150, | |
| "boolean:6-10": 150, | |
| "boolean:11-25": 150, | |
| "boolean:26-254": 150, | |
| "choice:2": 150, | |
| "choice:3-5": 150, | |
| "choice:6-10": 150, | |
| "choice:11-25": 150, | |
| "choice:26-254": 150, | |
| "ordinal:2": 150, | |
| "ordinal:3-5": 150, | |
| "ordinal:6-10": 150, | |
| "ordinal:11-25": 150, | |
| "ordinal:26-254": 150 | |
| } | |
| } |