Closed aojunzz closed 11 months ago
I try to convert the NeoX checkpoint to HF format using convert_sequential_to_hf.py, the config is https://github.com/EleutherAI/gpt-neox/blob/6bc724b90895fb7de7d324a385e5ca6992d54e9e/configs/llemma_7b.yml,
But, I encountered the error: KeyError: 'sequential.2.mlp.dense_4h_to_h.weight'
Hi, thanks for raising an issue!
It looks like I did not check in our updated-for-llama conversion scripts to our submodule branch--I will go do this.
I try to convert the NeoX checkpoint to HF format using convert_sequential_to_hf.py, the config is https://github.com/EleutherAI/gpt-neox/blob/6bc724b90895fb7de7d324a385e5ca6992d54e9e/configs/llemma_7b.yml,
But, I encountered the error: KeyError: 'sequential.2.mlp.dense_4h_to_h.weight'