CERC-AAI / multimodal

An implementation of model parallel autoregressive transformers on GPUs, based on the DeepSpeed library.
Apache License 2.0
8 stars 3 forks source link

Setup an interactive inference script on Colab to see how the new MAGMA checkpoints perform #25

Open kshitijkg opened 1 year ago

kshitijkg commented 1 year ago

1) Convert Robin Model to HF checkpoint. For this you need to extend the GPT-NeoX class in HF and add CLIP encoder, and adapter to it and adapt the conversion script by including clip and adapter weights 2) Load hf checkpoint in google colab 3) write boilerplate code for interacting with it

Alexis-BX commented 1 year ago

You say you want it on Jupyter, but how? To my knowledge we can't run jupyters on Summit. So where and how do you want to run this?

kshitijkg commented 1 year ago

27