Scandinavian XLM-RoBERTa (base-sized model)
This model is currently being created. Do not use yet.
Adjusting down lr from 1e4 to 5e5 since we have some instability. Restarting Nov 6. Training only 500k steps.
This model is currently being created. Do not use yet.
Adjusting down lr from 1e4 to 5e5 since we have some instability. Restarting Nov 6. Training only 500k steps.