Qwen 2.5 VL Training Plan?
I have question. Do you plan to train Qwen 2.5 VL? I think I will use it with Qwen-image.
Hi, currently working on new dataset, more versatile and diverse to achieve better functionality and accuracy. If qwen-3-vl or another really good model is not released when it's done, then the training will be based on qwen-2.5-vl.
I have a question regarding Qwen3-VL 8B/Qwen3-VL 30B: Do you think fine-tuning will be necessary?
Hi, soon I'm going to try with Qwen3-VL 4B. If it will be successful - there will be a version for 8b.
The dataset is mostly ready (main part is published). However, currently I can't name any dates or make promises when it's done.
Can you perform fine-tuning with video data? with What original model do you use, or is it an abliterated model?
To add some video data in dataset of extra finetune with it after main training - there are chances. As for base model - it is better to use original base without extra interventions, unless you know exactly what you want.