BAAI-DCAI/Bunny

Issues

• Can't Reproduce the result of the Bunny bsaed Phi-3

  #142 opened 13 days ago by bollossom
  1

• 导入checkpoint

  #141 opened 15 days ago by renyingxuan
  1

• when unfreeze_vision_tower set true No matter how the image changes, the model outputs the exact same answer

  #140 opened 22 days ago by Gary2018X
  3

• How to train Bunny with both image-text pair data and pure text data together?

  #139 opened a month ago by isbrycee
  2

• Question about deepspeed checkpoint loading

  #138 opened a month ago by Wintoplay
  1

• What is the text encoder for the bunny model？

  #131 opened a month ago by Why0912
  2

• Question: What is the minimum size of an image that can be classified?

  #128 opened a month ago by ChangGiMoon
  2

• evaluation about mmbench

  #125 opened a month ago by caiyuxuan1120
  2

• about images of pretrain data

  #124 opened a month ago by nicedoctor
  2

• Multi-images in 1 prompt

  #120 opened a month ago by motcapbovit
  2

• About pretrain data

  #119 opened a month ago by kid369
  5

• Convert Bunny-v1.0-3B to GGUF

  #115 opened a month ago by q104769424
  6

• Vision_tower is not updated as expected

  #130 opened a month ago by ChenFicha
  8

• About continuous_training

  #137 opened a month ago by Wintoplay
  3

• what is questions_answers_YN?

  #136 opened a month ago by Tramac
  1

• Trying to test checkpoint of a model during training process

  #135 opened a month ago by Wintoplay
  1

• Faster inference

  #134 opened 2 months ago by sahil02235
  0

• Inconsistency in Bunny-695K Dataset with Technical Report and Sample Duplication

  #132 opened 2 months ago by daybreaksly
  1

• Batch inference

  #93 opened 6 months ago by mtsysin
  5

• Can this model do object detection job?

  #129 opened 3 months ago by PredyDaddy
  0

• Any plan to integrate gemma-2-9b, llama-3.1-8b and phi-3.5

  #127 opened 3 months ago by sahilqure
  0

• obtain the log probabilities of output tokens

  #126 opened 4 months ago by Yuxin916
  2

• Continuous Fine-tuning Bunny 1.1 4B

  #123 opened 4 months ago by ChenFicha
  5

• Download the model to local inference, but still link to huggingface.co

  #118 opened 4 months ago by Kyousoso
  3

• Pre_train 和 SFT

  #109 opened 4 months ago by mynamelxy
  2

• KeyError: 'bunny-phi'

  #105 opened 4 months ago by jui0616
  3

• NotImplementedError: Cannot copy out of meta tensor; no data!

  #101 opened 4 months ago by A-Akhil
  2

• Model only responds with fine-tuned answers

  #97 opened 4 months ago by tamdan17
  2

• Bunny v1.1 Llama 3 8b GGUF support/release?

  #122 opened 4 months ago
  1

• Inference acceleration, can the trained model use some inference framework? The code comes from the llava architecture. Can it be integrated into inference frameworks such as sglang or lmdeploy similar to llava?

  #121 opened 4 months ago by zhangqingwu
  0

• Training the model throws an error after quantization

  #112 opened 5 months ago by dingtine
  1

• 修改或融合视觉模块

  #99 opened 5 months ago by Why0912
  4

• about the training strategy for Llama-3-8B

  #96 opened 5 months ago by Jancsi9981
  2

• Network Error due to High Traffic Error Message

  #114 opened 5 months ago by kyuewang17
  1

• S2-Wrapper Strategy Training Resulting in Tensor Shape Mismatch

  #110 opened 5 months ago by dingtine
  3

• tokenization mismatch when fine-tuning bunny-phi3

  #111 opened 5 months ago by simplelifetime
  7

• Encountered an error during model inference after using qwen2.

  #107 opened 5 months ago by dingtine
  2

• Question on the Role of Text Encoder in Bunny MLLM Architecture

  #108 opened 5 months ago by codefanw
  2

• Why do you modify the function `prepare_inputs_for_generation` of the LLM?

  #106 opened 5 months ago by linhaojia13
  2

• Smaller qwen2 model?

  #104 opened 6 months ago by R3xpook
  1

• about finetune train

  #90 opened 5 months ago by ZuyongWu
  2

• 微调数据集制作疑问

  #103 opened 6 months ago by chenzhu005774
  1

• about training

  #95 opened 6 months ago by Tengfei000
  6

• 微调报错

  #102 opened 6 months ago by chenzhu005774
  5

• 微调模型后启动web显示矩阵维度对不上

  #100 opened 6 months ago by htesd
  2

• Zero3 error for pretrain

  #98 opened 6 months ago by zhww
  4

• Support for Qwen2

  #94 opened 6 months ago by Gary2018X
  0

• convert raw data to training format

  #92 opened 6 months ago by acul3
  4

• why use deepspeed zero2 for pretrain but use zero3 for finetune?

  #91 opened 6 months ago by double-fire-0
  2

• question about s2

  #89 opened 7 months ago by zezeze97
  5