ImportError: cannot import name 'LlavaLlamaForCausalLM' from 'llava.model' (/root/LLaVA/llava/model/__init__.py)

[intelligence66]

Question

I encountered an issue where I was unable to import the model while running Lora fine-tuning again，this is the printed log:
Traceback (most recent call last):
File "/root/LLaVA/llava/train/train_mem.py", line 1, in
from llava.train.train import train
File "/root/LLaVA/llava/init.py", line 1, in
from .model import LlavaLlamaForCausalLM
ImportError: cannot import name 'LlavaLlamaForCausalLM' from 'llava.model' (/root/LLaVA/llava/model/init.py)

The script I am running is: sh ./scripts/v1_5/finetune_task_lora.sh

[SixCorePeach]

I also encount for this problem, is there lack of any file named the LlavaLlamaForCausalLM ？

[SixCorePeach]

oh, I see the init file,
AVAILABLE_MODELS = {
"llava_llama": "LlavaLlamaForCausalLM, LlavaConfig",
"llava_qwen": "LlavaQwenForCausalLM, LlavaQwenConfig",
"llava_mistral": "LlavaMistralForCausalLM, LlavaMistralConfig",
"llava_mixtral": "LlavaMixtralForCausalLM, LlavaMixtralConfig",
# "llava_qwen_moe": "LlavaQwenMoeForCausalLM, LlavaQwenMoeConfig",
# Add other models as needed
}
we could try to initial the corresponding model.
e.g. the llava_llama is in llava.model.language_model

when you solve this problem, there will get another one

[intelligence66]

There was a version conflict with the package, but I resolved it after reinstalling the package version.This is the version of the package I reinstalled:

[rajsinha02]

I am facing this exact same issue. Was anyone able to resolve successfully?

[liuxiang09]

这是 flash-attn 包与 PyTorch 版本不兼容导致的问题。首先pytorch版本要是12.1，然后如果你只是想先试一下demo，可以通过pip uninstall flash-attn ，卸载flash-attn。在其他条件正常的情况下，就不会出现这个报错了