---
license: apache-2.0
base_model:
- Qwen/Qwen2.5-VL-32B-Instruct
language:
- en
tags:
- BF16
- text-generation-inference
- uncensored
- abliterated
- unfiltered
- unredacted
- vllm
- pytorch
- max
- legal
pipeline_tag: image-text-to-text
library_name: transformers
---
 
![1](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/KwkklVMHdQMwxBJJ5pK5N.png)

# **Qwen2.5-VL-32B-Instruct-Unredacted-MAX**

> **Qwen2.5-VL-32B-Instruct-Unredacted-MAX** is an optimized release built on top of **huihui-ai/Qwen2.5-VL-32B-Instruct-abliterated**. This version focuses on **updated shard sizing, repository optimization, and compatibility improvements for the latest Transformers releases**, while preserving the multimodal reasoning and captioning capabilities of the original model. The result is a highly capable **32B vision-language model** designed for stable inference, efficient deployment, and modern ecosystem integration.

## Key Highlights

* **Optimized Repository Packaging**
  Improved model organization for smoother downloads, loading, and deployment workflows.

* **Latest Transformers Compatibility**
  Re-sharded and updated for improved compatibility with recent Transformers releases.

* **32B Vision-Language Architecture**
  Built on top of **Qwen2.5-VL-32B-Instruct**, delivering strong multimodal reasoning capacity.

* **Stable Multimodal Inference**
  Designed for consistent image and text processing across a range of deployment environments.

* **High-Fidelity Captioning**
  Suitable for detailed visual description, dataset generation, and multimodal analysis workflows.

* **Dynamic Resolution Support**
  Retains Qwen2.5-VL’s ability to handle varying image resolutions and aspect ratios effectively.

---

## Base Model Signatures:

This model has been re-sharded and optimized for the latest Transformers version from the base model: https://huggingface.co/huihui-ai/Qwen2.5-VL-32B-Instruct-abliterated.

---

## Quick Start with Transformers

```python
from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
import torch

model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    "prithivMLmods/Qwen2.5-VL-32B-Instruct-Unredacted-MAX",
    torch_dtype="auto",
    device_map="auto"
)

processor = AutoProcessor.from_pretrained(
    "prithivMLmods/Qwen2.5-VL-32B-Instruct-Unredacted-MAX"
)

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg",
            },
            {"type": "text", "text": "Provide a detailed caption for this image."},
        ],
    }
]

text = processor.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)

image_inputs, video_inputs = process_vision_info(messages)

inputs = processor(
    text=[text],
    images=image_inputs,
    videos=video_inputs,
    padding=True,
    return_tensors="pt",
).to("cuda")

generated_ids = model.generate(**inputs, max_new_tokens=256)

generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]

output_text = processor.batch_decode(
    generated_ids_trimmed,
    skip_special_tokens=True,
    clean_up_tokenization_spaces=False
)

print(output_text)
```

## Intended Use

* **Multimodal Research**
  Studying vision-language reasoning across image and text inputs.

* **Captioning and Dataset Work**
  Generating detailed descriptions for accessibility, annotation, and enrichment workflows.

* **Evaluation and Prototyping**
  Testing multimodal pipelines and experimenting with visual understanding tasks.

* **Local and High-Performance Deployment**
  Running large vision-language models on optimized GPU setups.

## Limitations & Risks

> **Important Note**: This model inherits the behavior and limitations of its base architecture.

* **Output Variability**
  Responses may vary depending on image quality, prompt design, and decoding settings.

* **Resource Requirements**
  The 32B model requires substantial VRAM, especially for high-resolution image processing.

* **Deployment Constraints**
  Performance depends on runtime optimization and hardware configuration.

* **General Model Limitations**
  May still produce incorrect, incomplete, or inconsistent outputs in complex scenarios.