luogen1996 LLaVA-HR issues

luogen1996 / LLaVA-HR

LLaVA-HR: High-Resolution Large Language-Vision Assistant

Apache License 2.0

192 stars 9 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Got loss 0 using ZeRO2 in sft stage

#16 Eddy-W closed 1 month ago
0
Error Loading Parameters for Slow Vision Tower During Fine-Tuning on Custom Dataset

#15 buckerF opened 2 months ago
0
Any plan to release your stage1 pretraining checkpoint?

#14 zjr2000 closed 2 months ago
0
【CUDA Version】

#13 Luo-Z13 opened 3 months ago
0
unfreeze vision encoder on SFT

#12 syp2ysy closed 3 months ago
0
locally downloaded model files doesn't seem to be supported by the code, as it always requires online downloading

#11 chuangzhidan opened 3 months ago
0
4bit support?

#10 chigkim opened 3 months ago
0
demo doesn't work : Error while finding module specification for 'llava.serve.cli' (ModuleNotFoundError: No module named 'llava')

#9 chuangzhidan opened 3 months ago
2
requests.exceptions.ConnectTimeout: (MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /timm/convnext_xxlarge.clip_laion2b_soup_ft_in1k/resolve/main/pytorch_model.bin (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x7f48e9d64670>, 'Connection to huggingface.co timed out. (connect timeout=10)'))"), '(Request ID: 52084a1b-5e0c-4d52-bcfb-41bf34c52df6)')

#8 gapjialin closed 4 months ago
14
How to load convnext model using local bin file

#7 yuecao0119 opened 4 months ago
6
Understanding how MR-Adapter works

#6 SoroushMehraban opened 4 months ago
0
Can we finetune LLaVA-HR with the scripts provided by the original LLaVA 1.5?

#5 zengxingchen opened 4 months ago
3
the training parameters of your single branch convnext encoder

#4 yuecao0119 opened 4 months ago
3
Got loss 0 after traing with larger input size in stage 2 (sft)

#3 luohao123 opened 4 months ago
4
Update README.md

#2 eltociear closed 4 months ago
0
Ablation study on using just single path encoder?

#1 lucasjinreal opened 4 months ago
9