issues
search
zjysteven
/
lmms-finetune
A minimal codebase for finetuning large multimodal models, supporting llava-1.5/1.6, llava-interleave, llava-next-video, qwen-vl, phi3-v etc.
Apache License 2.0
115
stars
8
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Any plans to support Qwen2-VL?
#36
KevinH48264
opened
14 hours ago
1
Is it possible to train just the projector not the LLM and the encoder?
#35
TonyJiang17
opened
2 days ago
7
Inference giving errors after finetuning LLavaInterleave 0.5B
#34
binarybeastt
opened
6 days ago
1
Trying to understand the Data Collator for compute_metrics evaluation
#33
KevinH48264
closed
6 days ago
3
Flash-attn issues
#32
binarybeastt
closed
6 days ago
2
Add gradio web
#31
zjysteven
closed
1 week ago
0
sync with main
#30
zjysteven
closed
1 week ago
0
How to support initializing training from a checkpoint?
#29
KevinH48264
closed
1 week ago
6
bigger model support
#28
gn64
opened
2 weeks ago
4
make it up-to-date
#27
zjysteven
closed
2 weeks ago
0
updates
#26
zjysteven
closed
2 weeks ago
0
LLaVA-OneVision for single-image, multi-image, and video
#25
KevinH48264
closed
2 weeks ago
2
How to integrate llava-next, and are there any specific things to pay attention to?
#24
Code-kunkun
closed
2 weeks ago
2
Let llava-interleave use assistant token mask when constructing labels
#23
zjysteven
closed
2 weeks ago
0
main -> chat-temp-new-models
#22
zjysteven
closed
2 weeks ago
0
error occurred: The input provided to the model are wrong. The number of image tokens is 4 while the number of image given to the model is 2. This prevents correct indexing and breaks batch generation.
#21
ReadyTeresa
closed
2 weeks ago
1
微调相关的问题
#20
evi-Genius
closed
2 weeks ago
3
move sampler to utils.py
#19
zjysteven
closed
4 weeks ago
0
Pre-compute vision encoder features
#18
nisargshah1999
closed
2 weeks ago
3
Full finetuning v.s. LoRA finetuning
#17
joeyy5588
closed
2 weeks ago
10
LLaVA-Next repo demo
#16
sarisel
closed
4 weeks ago
3
is it possible to support the fine-tuning of the llava-next-interleave series？
#15
qiuchen001
closed
4 weeks ago
10
Enabling finetuning of vision encoder and projector
#14
zjysteven
closed
1 month ago
0
Fixing "ValueError: No chat template is set for this processor."
#13
zjysteven
opened
1 month ago
0
Have you tested the llava-interleave-qwen-7b model on the ego4d_video_train.json dataset?
#12
Xavier-Zeng
closed
1 month ago
3
By default, it seems that learnable parameters are still added to the vision encoder?
#11
Code-kunkun
closed
1 month ago
8
[WIP] vision unfreeze
#10
linyueqian
closed
1 month ago
0
Phi-3-Vision训练出错
#9
yjmpepe
closed
1 month ago
2
运行example_image.sh报错
#8
yjmpepe
closed
1 month ago
10
working inference code for video model?
#7
Namzakku
closed
1 month ago
8
Avoid the List[np.ndarrys] -> torch.tensor warning
#6
zjysteven
closed
1 month ago
0
The "video" key is a must in llava_next_video in json data?
#5
Namzakku
closed
1 month ago
2
vision_feature_layer should be -2 not -1 in llava interleave
#4
sohananisetty
closed
1 month ago
7
Issue: Currently all vision modules are freezed for simplicity
#3
matbee-eth
closed
1 month ago
4
想要支持text-only数据需要做哪些修改呢
#2
fedshyvana
closed
1 month ago
4
作者您好，如果我想在slurm集群上提交finetune任务且我已经将llava-interleave-qwen-7b的参数下载到服务器上了，请问应该如何更改代码？
#1
whycantfindaname
closed
1 month ago
10