Passenger12138

I have implemented a basic functionality to visualize the attention maps of VDM (such as CogVideo) models. My first version of the code is available at [[my repository](https://github.com/Passenger12138/attention-map-diffusers-vdm.git)](https://github.com/Passenger12138/attention-map-diffusers-vdm.git). Currently, the...

我更新了一版基于cogvideo 1.5的lora微调代码，主要用于支持bucket训练并解决了cogvideo1.5位置编码和pfs 编码报错的问题 以下是我的代码https://github.com/Passenger12138/CogVideoX-5B-I2V-v1.5-lora-train.git

🚀 **We are excited to release the LoRA fine-tuning code for CogVideoX 1.5 by Diffusers**, designed specifically for image-to-video (image2video) tasks! This update brings significant improvements and new features to...

I’m fine-tuning **InternVL3_5-1B** (which uses a Qwen3-based LLM) with `transformers==4.57.3` and encountered the following warning: > *The tokenizer you are loading from '...' has an incorrect regex pattern [...] You...

Hi, @DN6, I tried the solution you suggested such as ``` # group offload apply_group_offloading( pipe.transformer, offload_type="leaf_level", offload_device=torch.device("cpu"), onload_device=torch.device("cuda"), use_stream=True, ) apply_group_offloading( pipe.transformer_2, offload_type="leaf_level", offload_device=torch.device("cpu"), onload_device=torch.device("cuda"), use_stream=True, ) apply_group_offloading( pipe.text_encoder,...