dvlab-research/MGM

Issues

• mgm-34b-hd, should have a 'model_type' key in its config.json

  #119 opened 10 months ago by chrisx599
  3

• 使用多gpu启动worker，对话时报错

  #125 opened 10 months ago by kimi360
  2

• Request for Access to Google's Gemma models family for Medical Science Research

  #139 opened 3 months ago by Geniusplug
  0

• Update LLaMA-3 tokenizer strategy

  #138 opened 4 months ago by ShaoTengLiu
  0

• error in loading

  #122 opened 10 months ago by TuuSiwei
  2

• lora initialisation missing from builder.py

  #106 opened a year ago by adrielkuek
  2

• comfyUI 开始任务出现如下错误

  #137 opened 5 months ago by jimmyyu1989
  0

• article error

  #136 opened 6 months ago by TuuSiwei
  0

• Unable to Merge LoRA Weights with Base Model: ValueError: Can't find 'adapter_config.json' at ...

  #135 opened 7 months ago by PARSA-MHMDI
  0

• TypeError: MGMllamaForCausalLM.forward() got an unexpected keyword argument 'cache_position'

  #133 opened 9 months ago by czm708033
  2

• CHAIR Evaluation

  #134 opened 8 months ago by itsqyh
  0

• dataset miss problem

  #116 opened 10 months ago by TuuSiwei
  1

• 使用cli调用自定义微调模型，出现'OpenCLIPVisionTower' object has no attribute 'device'

  #93 opened a year ago by HongLouyemeng
  16

• ImportError: cannot import name 'packaging' from 'pkg_resources'

  #132 opened 9 months ago by LiuRicky
  1

• when use stable-diffusion,AttributeError: 'NoneType' object has no attribute 'tokenize'

  #131 opened 9 months ago by ALR-alr
  1

• llama3 result is repeated many times

  #130 opened 10 months ago by pennypengpm
  1

• Does MGM support in-context(few-shot) inference?

  #129 opened 10 months ago by waltonfuture
  0

• Will there be support for Qwen2?

  #128 opened 10 months ago by huxian0402
  0

• How to access hidden states?

  #127 opened 10 months ago by Divyanshsingh1910
  1

• Error while loading model with transformers library

  #105 opened a year ago by PrateekPal641
  3

• I get this error: WARNING: tokenization mismatch: 156 vs. 161. (ignored) when I finetune llama3

  #126 opened 10 months ago by shidingz
  1

• May I ask if the current inference code does not support multi images input

  #114 opened a year ago by Angelalilyer
  1

• loss 0 and grad nan

  #123 opened 10 months ago by TuuSiwei
  3

• How to fix [NETWORK ERROR DUE TO HIGH TRAFFIC. ] on MacOS ?

  #99 opened a year ago by seasoncool
  2

• | EORROR | stderr | RecursionError: Maximun recursion depth exceeded in comparison

  #115 opened a year ago by linyf38
  1

• Do you meet the error "MGMConfig"?

  #124 opened 10 months ago by strawberryrs620
  0

• Requirement for pretraining weights of LLaMa-3-8B-Instruct

  #121 opened 10 months ago by shiwk23
  0

• Can provide laion-gpt4v dataset images zip?

  #120 opened 10 months ago by TuuSiwei
  0

• 关于多机多卡效果不如单机多卡好的问题

  #111 opened a year ago by DePengW
  1

• Inference problem about the demo.

  #118 opened 10 months ago by ApolloRay
  1

• The data for alignment and finetuning contains duplicates. Can you please explain why this is happening?

  #117 opened a year ago by KANGRuipeng
  1

• 可以放一下生成generation_pure_text数据的代码吗

  #109 opened a year ago by pennypengpm
  4

• Generation-related Instructions dataset link

  #112 opened a year ago by berry-ding
  1

• 多轮对话修改图像输入后报错

  #113 opened a year ago by pennypengpm
  0

• Loss does not decrease

  #110 opened a year ago by yfthu
  0

• pretrain error: lack of preprocessor_config.json

  #92 opened a year ago by jiezhangGt
  1

• Which deepspeed version is it

  #91 opened a year ago by Kareneveve
  2

• LLama 70B support

  #108 opened a year ago by PrateekPal641
  0

• Inference speed

  #107 opened a year ago by PrateekPal641
  0

• Some weights of the model checkpoint were not used when initializing MGMLlamaForCausalLM

  #103 opened a year ago by charlesCXK
  2

• how to use stage2 ckpt fine-tuning stage3？

  #102 opened a year ago by linqinguang
  1

• Excessive Length of Responses from Mini Gemini

  #97 opened a year ago by Dopplenum
  1

• Use of ocr in Evaluation

  #95 opened a year ago by bruceisme
  1

• Congratulations for the best LLaVA derived models !

  #104 opened a year ago by deepbeepmeep
  1

• 计划加入DPO训练来缓解模型幻觉问题吗

  #101 opened a year ago by jiezhangGt
  0

• AttributeError: 'OpenCLIPVisionTower' object has no attribute 'device'

  #96 opened a year ago by l1019008146
  2

• Take input image as condition.

  #100 opened a year ago by Adenialzz
  2

• stage2 loss is 0

  #98 opened a year ago by jiezhangGt
  1

• 'LlamaForCausalLM' object has no attribute 'get_vision_tower'

  #94 opened a year ago by HongLouyemeng
  1

• 当我使用推理命令的时候出现网络错误，无法构建推理的接口

  #90 opened a year ago by HongLouyemeng
  2