Zengjie Hu(胡增杰)

### Reminder - [x] I have read the above rules and searched the existing issues. ### System Info - `llamafactory` version: 0.9.3.dev0 - Platform: Linux-5.4.250-2-velinux1u1-amd64-x86_64-with-glibc2.31 - Python version: 3.10.16 -...

RuntimeError: CUDA driver error: invalid argument

9

### Reminder - [x] I have read the above rules and searched the existing issues. ### System Info - `llamafactory` version: 0.9.3.dev0 - Platform: Linux-5.4.250-2-velinux1u1-amd64-x86_64-with-glibc2.31 - Python version: 3.10.16 -...

Why does it take such a long time to perform SFT using LoRA?

16

### Reminder - [x] I have read the above rules and searched the existing issues. ### System Info - `llamafactory` version: 0.9.3.dev0 - Platform: Linux-5.4.250-2-velinux1u1-amd64-x86_64-with-glibc2.31 - Python version: 3.10.16 -...

Why does the loss suddenly drop to zero and the grad_norm stabilize at 1 at specific training steps when fine-tuning Llama3.2-vision-11B?

1

### Reminder - [x] I have read the above rules and searched the existing issues. ### System Info - `llamafactory` version: 0.9.3.dev0 - Platform: Linux-5.4.250-2-velinux1u1-amd64-x86_64-with-glibc2.31 - Python version: 3.10.16 -...

Error when excuting run.py with attack_strategy=escape/naive/ignore and defense=no

3

Hello, When I set attack_strategy=escape/naive/ignore and defense=no, the following error occurred,What could be the reason for this? @liu00222 ``` Traceback (most recent call last): File "/fs-computility/llmit_d/shared/huzengjie/safety_alignment/eval/Open-Prompt-Injection/main.py", line 74, in attack_responses...

How to obtain the ChartGemma test set you used? The paper mentions that you selected 1,000 data samples yourselves.