Oliver

输入图片尺寸高h和宽w不相等时报错Segmentation Fault

2

pytorch导出预训练resnet50模型，去掉fc层，输入设置为256x128，经过量化成uint8模型以后，发现在tm_classification_timvx基础上小修改后，-g 256,128时会出现Segmentation Fault，但是-g 256,256或者-g 128,128 却是能正常运行的，明明模型的输入设置是256x128，为什么经过resize的图片大小和模型设置不一致时能正常输出结果，一致的时候却会报错呢orz，下面是我的测试代码以及onnx和uint8模型，目前能够确认是在**run_graph**函数这里出错了 测试环境是khadas vim3 最新的npu驱动    链接：https://pan.baidu.com/s/1vnkseiQ3EeWeiXpwKfevRA 提取码：lq14 --来自百度网盘超级会员V4的分享  ```c++ /* * Licensed to the Apache Software Foundation (ASF) under one * or...

gather_norm issue

2

As the comments I write [here](https://github.com/huggingface/diffusers/pull/9340#issuecomment-2487662398), I'm wondering is the **gather_norm settings during training are identical to the configs given in this repo**? I see that CogVideoX-1.0(2B & 5B) is...

Inquiry about attention mask used for EAGLE-3 Training

3

May I ask why does the attention mask also need to be shifted during the training-time test process? From what I've read in the code, it seems the attention mask...