GSAI-ML

/

ReFusion

like 8

Follow

GSAI-ML 249

Hi there, thank you for the feedback. Could you please share the specific error logs you are encountering? Also, it would be helpful if you could provide the script you used to run the model.

first git clone https://github.com/ML-GSAI/ReFusion --recursive
then i edit in generate.py line 393 from 'cuda' to 'cpu'
then line 395 model_path="local model location" to my local model location
and run with python3 generate.py

$ python3 generate.py
/home/j/.local/lib/python3.11/site-packages/torch/cuda/init.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.
import pynvml # type: ignore[import]
Traceback (most recent call last):
File "/home/j/ReFusion/generate.py", line 12, in
from qwen3.modeling_qwen3_refusion import Qwen3ForCausalLM
File "/home/j/ReFusion/qwen3/modeling_qwen3_refusion.py", line 44, in
from transformers.utils import LossKwargs, auto_docstring, can_return_tuple, is_torch_flex_attn_available, logging
ImportError: cannot import name 'LossKwargs' from 'transformers.utils' (/home/j/.local/lib/python3.11/site-packages/transformers/utils/init.py)

this was fixed by qwen 3 max and then other errors.......

i dont have conda on my pc and no instructions in github README.md how to install,os is ubuntu 22.04

$ sudo apt install conda
[sudo] password for j:
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
E: Unable to locate package conda

i dont have conda on my pc and no instructions in github README.md how to install,os is ubuntu 22.04

$ sudo apt install conda
[sudo] password for j:
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
E: Unable to locate package conda

https://repo.anaconda.com/miniconda/
use AI to help you out.

may try thanks
but not very enthusiastic

Hi, it seems you do not have Conda installed correctly. The command sudo apt install conda typically does not work for this purpose.

I strongly recommend installing Miniconda first.
Please refer to the official installation guide here: https://docs.anaconda.com/miniconda/

Once you have installed Miniconda successfully, please strictly follow the commands provided in the Environment Setup section of our GitHub README to create and configure the environment.

may try thanks
but not very enthusiastic

really? go fuck yourself

Thanks for good words.

Thanks for good words.

I pointed you to the official Miniconda download because sudo apt install conda is never going to work . Instead of a simple 'thanks, I'll try that,' you whine about not being 'very enthusiastic.'
If you're waiting for someone to spoon-feed you step-by-step installation instructions for basic software in a model's issue tracker, you're in the wrong place. No one owes you enthusiasm, hand-holding, or custom tutorials here.
next time, try showing basic gratitude when someone takes the time to correct your mistake

@rzgar I also HATE ungratefulness! So I feel you friend! Really that's the kind of behavior that can tilt me for hours...
But yo, here I don't see what they did wrong, I personally simply understand it as a "thanks but unfortunately I don't know if I really feel like doing it..." in the sense of not being optimistic or maybe not having the motivation to do so... nothing toward you. And they said thanks! They never asked to be babysitted, instead they even tried to help ppl help them in the first place, sharing what they could..
That's just my vue but really I think they didn't deserve to be roasted :( Maybe there has been a misunderstanding
Good day all! Peace!

I just don't know how this guy is following 36 000 accounts xD but that's another story

I just don't know how this guy is following 36 000 accounts xD but that's another story

WT...., this is one of the craziest things I've seen in a while. lol, nice one! I honestly laughed for about a minute.
about my reply to him, honestly, I took a screenshot and gave it to my local AI (its personality is tuned by a system prompt written by grok) just copy-pasted it, and didn't even read it until he posted a reply :)

To be honest, I already knew their nickname because we I created my HF account like 1.5y ago, they followed me just few hours after creation. I always thought that was a bot... Until now!
Sorry but I didn't understand shit about the 2nd line of your answer :D but I guess nvm

"Who not work will not eat." - direct translation

"Who not work will not eat."

• direct translation

More like you've made one more mistake in your sentence

Had to check if I haven't drunk the wrong bottle cause this is getting quite psychedelic at this point lol - Please continue.

recomendation is removing all offtopic messages