2U1/Llama3.2-Vision-Finetune issues and pull requests

#27 - UserWarning: None of the inputs have requires_grad=True. Gradients will be None

Issue - State: open - Opened by gumona12 11 days ago - 1 comment

#26 - issue in data.py processor

Issue - State: open - Opened by sean0042 12 days ago - 1 comment

#25 - Issue in data.py

Issue - State: closed - Opened by Equijano24 13 days ago - 1 comment

#24 - Error related to finetune with lora vision

Issue - State: open - Opened by jinglun531 13 days ago - 1 comment

#23 - Training with text-only instruction data

Issue - State: open - Opened by sean0042 26 days ago - 1 comment

#22 - i try to start train with bash scripts/finetune.sh

Issue - State: open - Opened by MahmoudElsayedMahmoud about 1 month ago - 6 comments

#21 - Fine-tune for classification tasks

Issue - State: open - Opened by simoroma about 1 month ago - 5 comments

#20 - Is it possible to add prompt tuning related code? I.e. train additional input vectors

Issue - State: open - Opened by listen0425 about 2 months ago - 1 comment

#19 - UnboundLocalError: local variable 'inputs' referenced before assignment

Issue - State: closed - Opened by CynthiaChuang 2 months ago - 2 comments

#18 - Mixed modality training data

Issue - State: open - Opened by haon-chen 2 months ago - 10 comments

#17 - How to fine-tune LoRA of LLaMA-3.2 11B Vision using raw text dataset?

Issue - State: closed - Opened by yukiarimo 3 months ago - 2 comments

#16 - 可不可以支持alpaca格式数据集？

Issue - State: closed - Opened by Wang-Xujing 3 months ago - 1 comment

#15 - FP16 and deepspeed zero3 both break training in my testing ([rank0]: RuntimeError: weight should have at least three dimensions), this config works

Pull Request - State: open - Opened by Julian2002 3 months ago - 1 comment

#14 - Time out at NCCL work:909

Issue - State: closed - Opened by a370701524 3 months ago - 5 comments

#13 - Minimum GPU memory need for fine-tuning

Issue - State: open - Opened by alaminkawsar 3 months ago - 9 comments

#12 - Is it support pre-training?

Issue - State: closed - Opened by WangRongsheng 3 months ago - 1 comment

#11 - training hangs when <image> token is not first in the input sequence

Issue - State: closed - Opened by rothfels 4 months ago - 3 comments

#10 - Update environment.yaml

Pull Request - State: closed - Opened by rothfels 4 months ago

#9 - cannot create conda environment

Issue - State: closed - Opened by rothfels 4 months ago - 7 comments

#8 - fix chat template

Pull Request - State: closed - Opened by junhongwang418 4 months ago - 1 comment

#7 - RuntimeError: probability tensor contains either `inf`, `nan` or element < 0 for batch size=16

Issue - State: closed - Opened by NandhaKishorM 4 months ago - 1 comment

#6 - RuntimeError: The expanded size of the tensor (448) must match the existing size (446) at non-singleton dimension 2. Target sizes: [1, 32, 448, 38424]. Tensor sizes: [1, 1, 446, 38424]

Issue - State: closed - Opened by chuan573906361 4 months ago - 14 comments

#5 - preprocessor_config.json and tokenizer models are not saving after merging

Issue - State: closed - Opened by NandhaKishorM 4 months ago - 3 comments

#4 - vision model layer not found in the base model during merging

Issue - State: closed - Opened by NandhaKishorM 4 months ago - 8 comments

#3 - missing 1 required positional argument: 'processor'

Issue - State: closed - Opened by NandhaKishorM 4 months ago - 9 comments

#2 - Training time and Data format support

Issue - State: closed - Opened by LeslieTrue 4 months ago - 4 comments

#1 - KeyError: 'pixel_values'

Issue - State: closed - Opened by HMacro 4 months ago - 3 comments

Ecosyste.ms: Issues

GitHub / 2U1/Llama3.2-Vision-Finetune issues and pull requests