Is it possible to integrate qwen3vl into the model as CLIP?
#108
by q27916430 - opened
Is it possible to integrate qwen3vl into the model as CLIP? During my testing, I found that many issues encountered were due to unclear descriptions in qwen2.5vl. For example, a widely confusing incident involved the mix-up of male and female genitalia. This problem occurred precisely because qwen2.5VL uniformly referred to both male and female organs as "genitalia" in its descriptions, leading to such confusion.
q27916430 changed discussion status to closed