Rohit Gupta
Results
22
comments of
Rohit Gupta
@czczup I was able to fix this issue using latest version of `transformers` by adding `language_model.model.rotary_emb` to the device_map. I am using 2xH100 GPUs. I also had to move a...
@strand2013 the error message comes from the rotary embedding module, I checked if the weights for that module are placed on the right device in the device_map