dvlab-research/MGM Issues
为什么输出结果为nan呢
Closed 1llama3 result is repeated many times
Updated 1Will there be support for Qwen2?
UpdatedHow to access hidden states?
Updated 1使用多gpu启动worker,对话时报错
Closed 1loss 0 and grad nan
Closed 3dataset miss problem
Closed 1error in loading
Closed关于多机多卡效果不如单机多卡好的问题
Updated 1Inference problem about the demo.
Updated 1多轮对话修改图像输入后报错
UpdatedLoss does not decrease
ClosedWhich deepspeed version is it
Updated 2LLama 70B support
UpdatedInference speed
UpdatedUse of ocr in Evaluation
Updated 1请问为什么在训练llama的脚本中,预训练和微调所使用的conv不一样
Updated 1计划加入DPO训练来缓解模型幻觉问题吗
UpdatedSome questions about the demo
Updated 3Take input image as condition.
Closed 2stage2 loss is 0
Closed 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
Closed 2how to prompt to get short response
Updated 2Huggingface inference script
Updated 1Finetune
Updated 7