2U1 Qwen2-VL-Finetune issues - Githubissues

2U1 / Qwen2-VL-Finetune

An open-source implementaion for fine-tuning Qwen2-VL series by Alibaba Cloud.

Apache License 2.0

122 stars 13 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

RuntimeError: shape mismatch: value tensor of shape [49, 1536] cannot be broadcast to indexing result of shape [0, 1536]

#18 anshkumar closed 2 days ago
1
Mixed dataset

#17 univa-JASON opened 1 week ago
2
The training speed is slow

#16 kike-0304 opened 2 weeks ago
1
format object detection use Qwen2-VL-Finetune

#15 xjohnxjohn closed 2 weeks ago
1
Handled mixed modality

#14 mano3-1 opened 3 weeks ago
1
How does LoRA filter out the torch.nn.modules.Linear parameters in ViT?

#13 weilanzhikong opened 3 weeks ago
1
Shape Mismtach

#12 mano3-1 opened 3 weeks ago
8
Can it support data without images or videos?

#11 kike-0304 opened 4 weeks ago
4
qwen2-vl-7b lora sft; grad_norm: nan; why? learning rate 1e-5; and donot find the max_grad_norm

#10 weilanzhikong closed 3 weeks ago
2
Dataset Format Support

#9 yfvl closed 3 weeks ago
1
Noob question

#8 R3xpook closed 3 weeks ago
5
Script about qlora.

#7 yeppp27 closed 3 weeks ago
7
Inference script

#6 jetwu-create closed 1 month ago
4
how can i finetune this model with multi GPU?

#5 univa-JASON closed 1 month ago
2
Handle multiple datasets

#4 hongsamvo closed 2 months ago
2
Could you provide scripts finetune 70B model

#3 xiapengchng closed 1 month ago
4
The image_mask in the below image becomes all Falses, when i use LoRA.

#2 MaxKinny closed 2 months ago
4
Training video results in an error in data.py

#1 cedonley closed 1 month ago
6