license_name: qwen-research
license_link: https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct/blob/main/LICENSE
language:
- en
pipeline_tag: image-text-to-text
tags: - multimodal
- abliterated
- uncensored
library_name: transformers
base_model: - Qwen/Qwen2.5-VL-3B-Instruct
huihui-ai/Qwen2.5-VL-3B-Instruct-abliterated
This is an uncensored version of Qwen/Qwen2.5-VL-3B-Instruct created with abliteration (see remove-refusals-with-transformers to know more about it).
It was only the text part that was processed, not the image part.
ollama
You can use huihui_ai/qwen2.5-vl-abliterated:3b directly,
ollama run huihui_ai/qwen2.5-vl-abliterated:3b
GGUF
The official llama.cpp-b6907 has now been updated to support Qwen2.5-VL conversion to GGUF format and can be tested using llama-mtmd-cli.
The GGUF file has been uploaded.
llama-mtmd-cli -m huihui-ai/Qwen2.5-VL-3B-Instruct-abliterated/GGUF/ggml-model-Q4_K_M.gguf --mmproj huihui-ai/Qwen2.5-VL-3B-Instruct-abliterated/GGUF/mmproj-ggml-model-f16.gguf -c 4096 --image png/cc.jpg -p "Describe this image."
If it's just for chatting, you can use llama-cli.
llama-cli -m huihui-ai/Qwen2.5-VL-3B-Instruct-abliterated/GGUF/ggml-model-Q4_K_M.gguf -c 4096
Usage
You can use this model in your applications by loading it with Hugging Face's transformers library:
from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
"huihui-ai/Qwen2.5-VL-3B-Instruct-abliterated", torch_dtype="auto", device_map="auto"
)
processor = AutoProcessor.from_pretrained("huihui-ai/Qwen2.5-VL-3B-Instruct-abliterated")
image_path = "/tmp/test.png"
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"image": f"file://{image_path}",
},
{"type": "text", "text": "Describe 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",
)
inputs = inputs.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
)
output_text = output_text[0]
print(output_text)
Donation
Your donation helps us continue our further development and improvement, a cup of coffee can do it.
- bitcoin:
bc1qqnkhuchxw0zqjh2ku3lu4hq45hc6gy84uk70ge