deep_shf

> 做了一下调整可以跑通MFR测试，但是onnx_helper.py会报 > > ``` > 2023-03-31 18:13:30.296547356 [W:onnxruntime:, execution_frame.cc:835 VerifyOutputSizes] Expected shape from model of {1,512} does not match actual shape of {32,512} for output 1333 > err: max...

> 做了一下调整可以跑通MFR测试，但是onnx_helper.py会报 > > ``` > 2023-03-31 18:13:30.296547356 [W:onnxruntime:, execution_frame.cc:835 VerifyOutputSizes] Expected shape from model of {1,512} does not match actual shape of {32,512} for output 1333 > err: max...

> > 做了一下调整可以跑通MFR测试，但是onnx_helper.py会报 > > ``` > > 2023-03-31 18:13:30.296547356 [W:onnxruntime:, execution_frame.cc:835 VerifyOutputSizes] Expected shape from model of {1,512} does not match actual shape of {32,512} for output 1333 >...

> 请提供完整的报错信息和运行脚本 这是运行脚本： PRE_SEQ_LEN=128 LR=2e-2 NUM_GPUS=1 torchrun --standalone --nnodes=1 --nproc-per-node=$NUM_GPUS main.py \ --do_train \ --train_file AdvertiseGen/train.json \ --validation_file AdvertiseGen/dev.json \ --preprocessing_num_workers 10 \ --prompt_column content \ --response_column summary \ --overwrite_cache...

> 老师这个代码怎么改成可以实时显示 可以交流一下。

> 好的，这周日我整理一下自述文件。 感谢大佬,期待开源细节,我会快速用自己的数据进行实验,到时候再来请教您

> I've used a ready-made model by [Xiang Shin Wuu](https://github.com/xiang-wuu). You might take a look [here](https://blog.csdn.net/weixin_40293999/article/details/130977968) thanks, I got it done

import torch from modelscope import snapshot_download, AutoModel, AutoTokenizer torch.set_grad_enabled(False) # init model and tokenizer # model_dir = snapshot_download('Shanghai_AI_Laboratory/internlm-xcomposer2-vl-7b-4bit') model_dir = 'modelscope/internlm-xcomposer2-vl-7b' print(model_dir) # model_dir = 'Shanghai_AI_Laboratory/internlm-xcomposer2-vl-7b-4bit' print(model_dir) # ssss model...

> > 用这个modelscope的代码时 ，用int4版本会出现缺少配置文件的错误 OSError: /home/shf/.cache/modelscope/hub/Shanghai_AI_Laboratory/internlm-xcomposer2-vl-7b-4bit does not appear to have a file named config.json. Checkout 'https://huggingface.co//home/shf/.cache/modelscope/hub/Shanghai_AI_Laboratory/internlm-xcomposer2-vl-7b-4bit/None' for available files. > > @panzhang0212 作者能解答一下吗 稀里糊涂的就成功了 我用了transformer的推理 也用了modelscope的推理 都成了

> Thanks for sharing! I'd like to do some research based on this library. Compared with the original paper, what modules do we need to complete this model, in addition...