WoWolf/Qwen2_5vl-7b-fm-tuned
WoWolf/Qwen2_5vl-7b-fm-tuned is a 7 billion parameter vision-language model, fine-tuned from Qwen/Qwen2.5-VL-7B-Instruct. It was trained on a combination of mllm_demo, identity, and alpaca_en_demo datasets, leveraging a 32768 token context length. This model is designed for multimodal tasks, building upon the Qwen2.5-VL architecture to enhance its understanding and generation capabilities across image and text inputs.
Loading preview...
Overview
WoWolf/Qwen2_5vl-7b-fm-tuned is a 7 billion parameter multimodal model, specifically a fine-tuned version of the Qwen/Qwen2.5-VL-7B-Instruct base model. This model integrates vision and language capabilities, allowing it to process and understand both image and text inputs. It was trained using a learning rate of 1e-05 over 3 epochs with a total batch size of 16, utilizing a cosine learning rate scheduler with a 0.1 warmup ratio.
Key Capabilities
- Multimodal Understanding: Inherits and refines the vision-language capabilities of the Qwen2.5-VL architecture.
- Instruction Following: Benefits from fine-tuning on instruction-based datasets like alpaca_en_demo, enhancing its ability to follow user commands.
- Contextual Processing: Supports a substantial context length of 32768 tokens, enabling it to handle longer and more complex multimodal inputs.
Good for
- Applications requiring joint understanding of images and text.
- Tasks that benefit from instruction-tuned models in a multimodal context.
- Research and development in multimodal AI, particularly for exploring fine-tuning effects on Qwen2.5-VL models.