Yijiang Kris Liu

For ResNet in cifar experiment , the shortcut/downsampling is binarized to [-1,+1] via BinarizeConv2d. However in ImageNet experiment , the shortcut/downsampling remains the fullprecision approach via nn.Conv2d. Is this the...

Any plan of supporting larger LLAMA models ?

2

It seems like that the current script only support the smallest (7B) LLAMA model. Really expect to see extending to larger models.

no GPU usage and only CPU running when inference

2

Training operates well with GPUs. However when I run inference, no GPU usage and only CPU running... I use following script to do inference. ```bash python3 scripts/generate_lm.py --load_model_path models/llama-7b.bin --spm_model_path...

How to organize customized text dataset ?

1

I would like to express my gratitude for your hard work on the project. I came across a training script where you used the following code: ```python DATA_FILE_PATH = 'elon_musk_tweets.csv'...

Training leads to OOM even with an 80GB GPU card. Would you please give some advices ? ```bash ***** Running training ***** Num examples = 1799 Num Epochs = 1...