Image-Text-to-Text
Transformers
Safetensors
PyTorch
English
qwen3_5
text-generation-inference
uncensored
abliterated
unfiltered
unredacted
refusal-ablated
vllm
bf16
max
alignment-modified
reasoning
conversational
Eval Results (legacy)
Instructions to use prithivMLmods/Qwen3.5-9B-Unredacted-MAX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use prithivMLmods/Qwen3.5-9B-Unredacted-MAX with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="prithivMLmods/Qwen3.5-9B-Unredacted-MAX") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("prithivMLmods/Qwen3.5-9B-Unredacted-MAX") model = AutoModelForMultimodalLM.from_pretrained("prithivMLmods/Qwen3.5-9B-Unredacted-MAX", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use prithivMLmods/Qwen3.5-9B-Unredacted-MAX with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "prithivMLmods/Qwen3.5-9B-Unredacted-MAX" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "prithivMLmods/Qwen3.5-9B-Unredacted-MAX", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/prithivMLmods/Qwen3.5-9B-Unredacted-MAX
- SGLang
How to use prithivMLmods/Qwen3.5-9B-Unredacted-MAX 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 "prithivMLmods/Qwen3.5-9B-Unredacted-MAX" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "prithivMLmods/Qwen3.5-9B-Unredacted-MAX", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'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 "prithivMLmods/Qwen3.5-9B-Unredacted-MAX" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "prithivMLmods/Qwen3.5-9B-Unredacted-MAX", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use prithivMLmods/Qwen3.5-9B-Unredacted-MAX with Docker Model Runner:
docker model run hf.co/prithivMLmods/Qwen3.5-9B-Unredacted-MAX
Update README.md
Browse files
README.md
CHANGED
|
@@ -34,29 +34,30 @@ model-index:
|
|
| 34 |
|
| 35 |
# **Qwen3.5-9B-Unredacted-MAX**
|
| 36 |
|
| 37 |
-
> **Qwen3.5-9B-Unredacted-MAX** is an
|
| 38 |
|
| 39 |
> [!IMPORTANT]
|
| 40 |
-
> This model is
|
| 41 |
|
| 42 |
---
|
| 43 |
|
| 44 |
## Base Model Signatures:
|
| 45 |
|
| 46 |
-
This model has been re-sharded and optimized for the latest Transformers version from the base model:
|
|
|
|
| 47 |
|
| 48 |
---
|
| 49 |
|
| 50 |
## Evaluation Report (Self-Reported)
|
| 51 |
|
| 52 |
-
**Model:**
|
| 53 |
|
| 54 |
-
|
| 55 |
-
|
| 56 |
|
| 57 |
-
> The evaluation was conducted using **2000
|
| 58 |
|
| 59 |
-
##
|
| 60 |
|
| 61 |
```yaml
|
| 62 |
evaluation:
|
|
@@ -64,7 +65,7 @@ evaluation:
|
|
| 64 |
total_test_prompts: 2000
|
| 65 |
evaluation_runs: 10
|
| 66 |
prompts_per_run: 200
|
| 67 |
-
evaluation_type:
|
| 68 |
|
| 69 |
results:
|
| 70 |
refusal_rate: 5.500
|
|
@@ -72,22 +73,35 @@ results:
|
|
| 72 |
abliteration_rate: 94.500
|
| 73 |
```
|
| 74 |
|
| 75 |
-
> Note:
|
| 76 |
|
| 77 |
---
|
| 78 |
|
| 79 |
## Key Highlights
|
| 80 |
|
| 81 |
-
* **
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
* **Improved
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
|
| 87 |
## Quick Start with Transformers
|
| 88 |
|
| 89 |
-
```
|
| 90 |
-
pip install transformers==5.3.0
|
|
|
|
|
|
|
| 91 |
```
|
| 92 |
|
| 93 |
```python
|
|
@@ -125,12 +139,8 @@ inputs = processor(
|
|
| 125 |
|
| 126 |
generated_ids = model.generate(**inputs, max_new_tokens=256)
|
| 127 |
|
| 128 |
-
generated_ids_trimmed = [
|
| 129 |
-
out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
|
| 130 |
-
]
|
| 131 |
-
|
| 132 |
output_text = processor.batch_decode(
|
| 133 |
-
|
| 134 |
skip_special_tokens=True,
|
| 135 |
clean_up_tokenization_spaces=False
|
| 136 |
)
|
|
@@ -138,24 +148,22 @@ output_text = processor.batch_decode(
|
|
| 138 |
print(output_text)
|
| 139 |
```
|
| 140 |
|
|
|
|
|
|
|
| 141 |
## Intended Use
|
| 142 |
|
| 143 |
-
*
|
| 144 |
-
*
|
| 145 |
-
*
|
| 146 |
-
*
|
| 147 |
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
> **Important Note**: This model intentionally reduces built-in refusal mechanisms.
|
| 151 |
|
| 152 |
-
|
| 153 |
-
* **User Responsibility**: Outputs should be handled responsibly and within legal and ethical boundaries.
|
| 154 |
-
* **Model Size Constraints**: Although stronger than smaller variants, a **9B model** still has limitations compared to very large frontier architectures.
|
| 155 |
|
| 156 |
-
|
| 157 |
|
| 158 |
-
*
|
| 159 |
-
*
|
| 160 |
-
*
|
| 161 |
-
*
|
|
|
|
| 34 |
|
| 35 |
# **Qwen3.5-9B-Unredacted-MAX**
|
| 36 |
|
| 37 |
+
> **Qwen3.5-9B-Unredacted-MAX** is an optimized release built on top of **huihui-ai/Huihui-Qwen3.5-9B-abliterated**. This version focuses on **improved packaging, updated repository structure, and enhanced compatibility with modern Transformers pipelines**, while preserving the reasoning and instruction-following behavior of the base model. The result is a capable **9B parameter language model** designed for efficient deployment, stable inference, and research-oriented experimentation.
|
| 38 |
|
| 39 |
> [!IMPORTANT]
|
| 40 |
+
> This model is intended for research and learning purposes only. Any outputs generated by this model are the sole responsibility of the user. The authors and hosting platform disclaim all liability for generated content. Users must ensure safe, ethical, and lawful usage.
|
| 41 |
|
| 42 |
---
|
| 43 |
|
| 44 |
## Base Model Signatures:
|
| 45 |
|
| 46 |
+
This model has been re-sharded and optimized for the latest Transformers version from the base model:
|
| 47 |
+
https://huggingface.co/huihui-ai/Huihui-Qwen3.5-9B-abliterated
|
| 48 |
|
| 49 |
---
|
| 50 |
|
| 51 |
## Evaluation Report (Self-Reported)
|
| 52 |
|
| 53 |
+
**Model:** Qwen3.5-9B-Unredacted-MAX
|
| 54 |
|
| 55 |
+
* **Abliteration Rate (Non-Refusal Rate):** 94.500
|
| 56 |
+
* **Refusal Rate:** 5.500
|
| 57 |
|
| 58 |
+
> The evaluation was conducted using **2000 test prompts** across multiple evaluation runs to measure model response behavior. Results are averaged and may vary depending on benchmarking setup, sampling strategy, and prompt distribution.
|
| 59 |
|
| 60 |
+
### Evaluation Summary (YAML)
|
| 61 |
|
| 62 |
```yaml
|
| 63 |
evaluation:
|
|
|
|
| 65 |
total_test_prompts: 2000
|
| 66 |
evaluation_runs: 10
|
| 67 |
prompts_per_run: 200
|
| 68 |
+
evaluation_type: response_behavior_analysis
|
| 69 |
|
| 70 |
results:
|
| 71 |
refusal_rate: 5.500
|
|
|
|
| 73 |
abliteration_rate: 94.500
|
| 74 |
```
|
| 75 |
|
| 76 |
+
> Note: These results are self-reported and should be interpreted as approximate indicators of behavior rather than strict benchmarks.
|
| 77 |
|
| 78 |
---
|
| 79 |
|
| 80 |
## Key Highlights
|
| 81 |
|
| 82 |
+
* **Optimized Repository Structure**
|
| 83 |
+
Streamlined model packaging for easier loading and deployment.
|
| 84 |
+
|
| 85 |
+
* **Improved Transformer Compatibility**
|
| 86 |
+
Designed to work smoothly with modern Hugging Face Transformers versions.
|
| 87 |
+
|
| 88 |
+
* **9B Parameter Architecture**
|
| 89 |
+
Built on **Qwen3.5-9B**, balancing performance and efficiency.
|
| 90 |
+
|
| 91 |
+
* **Stable Instruction Following**
|
| 92 |
+
Maintains consistent behavior across structured and multi-step prompts.
|
| 93 |
+
|
| 94 |
+
* **Efficient Deployment**
|
| 95 |
+
Suitable for local inference, prototyping, and research workflows.
|
| 96 |
+
|
| 97 |
+
---
|
| 98 |
|
| 99 |
## Quick Start with Transformers
|
| 100 |
|
| 101 |
+
```bash
|
| 102 |
+
pip install transformers==5.3.0
|
| 103 |
+
# or
|
| 104 |
+
pip install git+https://github.com/huggingface/transformers.git
|
| 105 |
```
|
| 106 |
|
| 107 |
```python
|
|
|
|
| 139 |
|
| 140 |
generated_ids = model.generate(**inputs, max_new_tokens=256)
|
| 141 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 142 |
output_text = processor.batch_decode(
|
| 143 |
+
[out[len(inp):] for inp, out in zip(inputs.input_ids, generated_ids)],
|
| 144 |
skip_special_tokens=True,
|
| 145 |
clean_up_tokenization_spaces=False
|
| 146 |
)
|
|
|
|
| 148 |
print(output_text)
|
| 149 |
```
|
| 150 |
|
| 151 |
+
---
|
| 152 |
+
|
| 153 |
## Intended Use
|
| 154 |
|
| 155 |
+
* Research into transformer behavior and instruction following
|
| 156 |
+
* Red-teaming and robustness evaluation
|
| 157 |
+
* Local and cloud inference deployment
|
| 158 |
+
* Rapid prototyping of NLP applications
|
| 159 |
|
| 160 |
+
---
|
|
|
|
|
|
|
| 161 |
|
| 162 |
+
## Limitations & Risks
|
|
|
|
|
|
|
| 163 |
|
| 164 |
+
> **Important Note**: This model inherits limitations from its base architecture.
|
| 165 |
|
| 166 |
+
* Outputs may vary depending on decoding settings and prompts
|
| 167 |
+
* Requires GPU acceleration for optimal performance
|
| 168 |
+
* May produce incorrect or inconsistent information in complex scenarios
|
| 169 |
+
* Performance is dependent on deployment environment and optimization strategy
|